{"id":"153cc8721a71d0d17f78645be14915dd77b0b082ad13f374bc98491a188ee7d7","title":"Hugging Face article presents Reson8’s Rumble-2-Flash speech model","argument":"Hugging Face’s article announces Reson8’s open speech recognition model Rumble-2-Flash, with 0.5B parameters and support for more than 40 languages. Reson8 says it competes with models 3–60x its size and runs in Transformers and vLLM without custom code. The team says it used on-policy distillation to make a smaller model: the student transcribes audio, then learns from a larger model’s token-by-token feedback. GOLD handles differences between the models’ tokenizers. On the Open ASR Leaderboard multilingual track, Reson8 reports a 6.96% average error rate across nine languages; on the English track, it reports 6.6% WER. On AppTek’s benchmark of 302 hours of helpline calls in 14 English accents, the model scored 7.86% WER, compared with 7.78% for NVIDIA’s Nemotron 3.5 Transcribe. Reson8 also reports throughput above 7,000 RTFx on larger leaderboard datasets with full batches, equivalent to about half a second for an hour of audio. The article presents these as benchmark results; it does not establish performance on other audio or in other deployment settings.","topicIds":["intelligence-and-discovery","work-and-agency"],"domain":"work","claimType":"attributed reading","limits":"The performance figures are claims reported in the post and are tied to the named benchmarks and conditions; they do not establish performance on other audio or in other deployment settings. The post reports throughput for larger leaderboard datasets where batches stay full, so that figure is not presented as a result for every workload. ","speaker":{"name":"Reson8, Mark 't Hart, Eustache Le Bihan, Eric Bezzam","roleAtSource":"Article author"},"source":{"title":"Rumble-2-Flash: an open 0.5B speech model for 40+ languages","publisher":"Hugging Face","format":"Article","url":"https://huggingface.co/blog/rumble-2-flash","publishedAt":"2026-11-18T00:00:00.772Z","dateBasis":"Date stated by the publisher."},"evidence":[{"text":"Reson8 announced Rumble-2-Flash as an open speech recognition model with 0.5B parameters and support for more than 40 languages.","excerpt":"Today we're open sourcing Rumble-2-Flash under the Apache 2.0 license. It's a 0.5B-parameter speech recognition model for 40+ languages","url":"https://huggingface.co/blog/rumble-2-flash","relationship":"Source-stated"},{"text":"Reson8 says the model was trained with on-policy distillation, in which the student transcribes audio and receives token-by-token feedback from the teacher.","excerpt":"The student doesn't train on transcripts the teacher wrote. It transcribes the audio itself, and the teacher grades that transcript token by token","url":"https://huggingface.co/blog/rumble-2-flash","relationship":"Source-stated"},{"text":"The post reports a 6.96% average error rate on the Open ASR Leaderboard multilingual track across nine named languages, and 6.6% WER on its English track.","excerpt":"Rumble-2-Flash averages a 6.96% error rate across Chinese, German, French, Italian, Spanish, Portuguese, Hindi, Dutch and English. That's better than Meta's omniASR-LLM 7B","url":"https://huggingface.co/blog/rumble-2-flash","relationship":"Source-stated"},{"text":"On AppTek’s call-centre benchmark, Reson8 reports 7.86% WER for Rumble-2-Flash and 7.78% for NVIDIA’s Nemotron 3.5 Transcribe.","excerpt":"Rumble-2-Flash reaches 7.86% WER, on par with NVIDIA's Nemotron 3.5 Transcribe (30B) at 7.78%.","url":"https://huggingface.co/blog/rumble-2-flash","relationship":"Source-stated"},{"text":"Reson8 reports throughput above 7,000 RTFx on larger leaderboard datasets when batches stay full.","excerpt":"on the larger datasets in HF's Open ASR Leaderboard, where batches stay full, it reaches over 7,000 RTFx.","url":"https://huggingface.co/blog/rumble-2-flash","relationship":"Source-stated"}],"sourceMedia":{"status":"found","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","checkedAt":"2026-11-18T19:50:11.127Z","items":[],"note":"Media was declared by the source, but no asset passed the supported availability checks. Open the original source to view it."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How does Rumble-2-Flash perform on audio and in deployment settings beyond the benchmarks described in the article?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Hugging Face article presents Reson8’s Rumble-2-Flash speech model","summary":"Hugging Face’s article announces Reson8’s open speech recognition model Rumble-2-Flash, with 0.5B parameters and support for more than 40 languages. Reson8 says it competes with models 3–60x its size and runs in Transformers and vLLM without custom code. The team says it used on-policy distillation to make a smaller model: the student transcribes audio, then learns from a larger model’s token-by-token feedback. GOLD handles differences between the models’ tokenizers. On the Open ASR Leaderboard multilingual track, Reson8 reports a 6.96% average error rate across nine languages; on the English track, it reports 6.6% WER. On AppTek’s benchmark of 302 hours of helpline calls in 14 English accents, the model scored 7.86% WER, compared with 7.78% for NVIDIA’s Nemotron 3.5 Transcribe. Reson8 also reports throughput above 7,000 RTFx on larger leaderboard datasets with full batches, equivalent to about half a second for an hour of audio. The article presents these as benchmark results; it does not establish performance on other audio or in other deployment settings.","question":"How does Rumble-2-Flash perform on audio and in deployment settings beyond the benchmarks described in the article?","claims":[{"text":"Reson8 announced Rumble-2-Flash as an open speech recognition model with 0.5B parameters and support for more than 40 languages.","excerpt":"Today we're open sourcing Rumble-2-Flash under the Apache 2.0 license. It's a 0.5B-parameter speech recognition model for 40+ languages","support":"source-stated"},{"text":"Reson8 says the model was trained with on-policy distillation, in which the student transcribes audio and receives token-by-token feedback from the teacher.","excerpt":"The student doesn't train on transcripts the teacher wrote. It transcribes the audio itself, and the teacher grades that transcript token by token","support":"source-stated"},{"text":"The post reports a 6.96% average error rate on the Open ASR Leaderboard multilingual track across nine named languages, and 6.6% WER on its English track.","excerpt":"Rumble-2-Flash averages a 6.96% error rate across Chinese, German, French, Italian, Spanish, Portuguese, Hindi, Dutch and English. That's better than Meta's omniASR-LLM 7B","support":"source-stated"},{"text":"On AppTek’s call-centre benchmark, Reson8 reports 7.86% WER for Rumble-2-Flash and 7.78% for NVIDIA’s Nemotron 3.5 Transcribe.","excerpt":"Rumble-2-Flash reaches 7.86% WER, on par with NVIDIA's Nemotron 3.5 Transcribe (30B) at 7.78%.","support":"source-stated"},{"text":"Reson8 reports throughput above 7,000 RTFx on larger leaderboard datasets when batches stay full.","excerpt":"on the larger datasets in HF's Open ASR Leaderboard, where batches stay full, it reaches over 7,000 RTFx.","support":"source-stated"}],"limitations":["The performance figures are claims reported in the post and are tied to the named benchmarks and conditions; they do not establish performance on other audio or in other deployment settings.","The post reports throughput for larger leaderboard datasets where batches stay full, so that figure is not presented as a result for every workload. "],"mediaId":"5b89a2db7f60ae9099b72ca73768977f61c082125cbaf94b72c9c9bced5f17aa","correctionReason":null,"id":"153cc8721a71d0d17f78645be14915dd77b0b082ad13f374bc98491a188ee7d7","publicationId":"c5432fb2-f7f2-4d3e-8577-011c3a71fcc4","version":1,"sourceUrl":"https://huggingface.co/blog/rumble-2-flash","sourceTitle":"Rumble-2-Flash: an open 0.5B speech model for 40+ languages","publisher":"Hugging Face","publishedAt":"2026-11-18T00:00:00.772Z","originalPublished":{"value":"2026-11-18T00:00:00.772Z","precision":"instant","basis":"publisher_page"},"firstObservedAt":"2026-11-18T14:43:04.732Z","createdAt":"2026-11-18T19:34:19.294Z","historical":false,"entityIds":["hugging-face"],"topicIds":["intelligence-and-discovery","work-and-agency"],"media":[{"id":"5b89a2db7f60ae9099b72ca73768977f61c082125cbaf94b72c9c9bced5f17aa","kind":"article","url":"https://huggingface.co/blog/rumble-2-flash","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-18T19:33:08.199Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"found","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","checkedAt":"2026-11-18T19:50:11.127Z","items":[],"note":"Media was declared by the source, but no asset passed the supported availability checks. Open the original source to view it."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v9","mediaId":"5b89a2db7f60ae9099b72ca73768977f61c082125cbaf94b72c9c9bced5f17aa","confidence":0.53,"provider":"typesafe","model":"jev-1.13.0","inputHash":"09dc399502e8d1666f7e258f14a019dfc71a19db846889a7a7c5c93c4fa0d10c","decidedAt":"2026-11-18T19:33:11.257Z"},"provenance":{"textHash":"534bf6bbac111d013823ea644911face72ac5c197e68c86057de56a40d213234","observationId":"9af96bcd1e8c2caa39933ac19f0102ef7576081e66b20658eb6de013b827ff6c","retrievedAt":"2026-11-18T19:33:07.792Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"553f8618-397a-4779-8663-a26a18427f1e"},"reviewer":{"model":"gpt-6-luna","runId":"1474d691-c893-4fe5-89e3-6358db91d3cf"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-18T19:34:19.294Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Reson8","url":null,"kind":"Person"},{"name":"Mark 't Hart","url":null,"kind":"Person"},{"name":"Eustache Le Bihan","url":"https://huggingface.co/eustlb","kind":"Person"},{"name":"Eric Bezzam","url":"https://huggingface.co/bezzam","kind":"Person"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Reson8","url":null,"kind":"Person","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Reson8","status":"stated","provenance":{"basis":"article-jsonld","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","locator":"document/html[1]/head[0]/script[43]@3226.author"},"assertionId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-18T19:50:11.127Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","contentSha256":"2afea669ef2d2382c1c089cabf7466f418cb86c370f51a872e53b09bac276675","textHash":"534bf6bbac111d013823ea644911face72ac5c197e68c86057de56a40d213234","observedAt":"2026-11-18T19:50:11.127Z"}},{"actor":{"name":"Mark 't Hart","url":null,"kind":"Person","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Mark 't Hart","status":"stated","provenance":{"basis":"article-jsonld","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","locator":"document/html[1]/head[0]/script[43]@3226.author"},"assertionId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-18T19:50:11.127Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","contentSha256":"2afea669ef2d2382c1c089cabf7466f418cb86c370f51a872e53b09bac276675","textHash":"534bf6bbac111d013823ea644911face72ac5c197e68c86057de56a40d213234","observedAt":"2026-11-18T19:50:11.127Z"}},{"actor":{"name":"Eustache Le Bihan","url":"https://huggingface.co/eustlb","kind":"Person","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Eustache Le Bihan","status":"stated","provenance":{"basis":"article-jsonld","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","locator":"document/html[1]/head[0]/script[43]@3226.author"},"assertionId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-18T19:50:11.127Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","contentSha256":"2afea669ef2d2382c1c089cabf7466f418cb86c370f51a872e53b09bac276675","textHash":"534bf6bbac111d013823ea644911face72ac5c197e68c86057de56a40d213234","observedAt":"2026-11-18T19:50:11.127Z"}},{"actor":{"name":"Eric Bezzam","url":"https://huggingface.co/bezzam","kind":"Person","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Eric Bezzam","status":"stated","provenance":{"basis":"article-jsonld","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","locator":"document/html[1]/head[0]/script[43]@3226.author"},"assertionId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-18T19:50:11.127Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","contentSha256":"2afea669ef2d2382c1c089cabf7466f418cb86c370f51a872e53b09bac276675","textHash":"534bf6bbac111d013823ea644911face72ac5c197e68c86057de56a40d213234","observedAt":"2026-11-18T19:50:11.127Z"}},{"actor":{"name":"Hugging Face","url":"https://huggingface.co/","kind":"unknown","actorId":null},"role":"publisher","scope":{"kind":"publication","locator":null},"literalCredit":"Hugging Face","status":"stated","provenance":{"basis":"article-publisher","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","locator":"document/html[1]/head[0]/script[43]@3226.publisher"},"assertionId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-18T19:50:11.127Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","contentSha256":"2afea669ef2d2382c1c089cabf7466f418cb86c370f51a872e53b09bac276675","textHash":"534bf6bbac111d013823ea644911face72ac5c197e68c86057de56a40d213234","observedAt":"2026-11-18T19:50:11.127Z"}}],"participationCoverage":{"total":5,"returned":5,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":{"name":"Hugging Face","url":"https://huggingface.co/"},"sourceUrl":"https://huggingface.co/blog/rumble-2-flash","snapshotId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00","contentSha256":"2afea669ef2d2382c1c089cabf7466f418cb86c370f51a872e53b09bac276675","textHash":"534bf6bbac111d013823ea644911face72ac5c197e68c86057de56a40d213234","observedAt":"2026-11-18T19:50:11.127Z","basis":"publisher-page-metadata","sourceLinks":50,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/153cc8721a71d0d17f78645be14915dd77b0b082ad13f374bc98491a188ee7d7/source"}},"canonicalUrl":"https://escapemeridian.com/readings/hugging-face-article-presents-reson8-s-rumble-2-flash-speech-153cc872","formats":{"html":"/readings/hugging-face-article-presents-reson8-s-rumble-2-flash-speech-153cc872","markdown":"/readings/hugging-face-article-presents-reson8-s-rumble-2-flash-speech-153cc872.md","json":"/readings/hugging-face-article-presents-reson8-s-rumble-2-flash-speech-153cc872.json"},"attribution":"Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.","maintainedGuideLinks":[],"conditions":[],"readerContext":{"version":"reader-context-v1","attribution":"Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.","terms":[{"id":"term-tokens","name":"Model tokens","explanation":"Small chunks of text, often parts of words, that a language model reads and writes; usage and prices are often counted in tokens.","referenceUrl":"https://developers.google.com/machine-learning/glossary#token"},{"id":"term-benchmark","name":"Benchmark","explanation":"A defined set of tests used to compare systems. Results depend on the tasks and measures selected.","referenceUrl":"https://arxiv.org/abs/2211.09110"}],"entities":[{"id":"hugging-face","name":"Hugging Face","kind":"organization","context":"Hugging Face hosts models, datasets and software that let people share, adapt and inspect AI systems.","referenceUrls":["https://huggingface.co/blog"],"path":"/entities/hugging-face","json":"/entities/hugging-face.json","markdown":"/entities/hugging-face.md"},{"id":"nvidia","name":"NVIDIA","kind":"organization","context":"NVIDIA develops processors, networking and software used for AI, graphics and scientific computing.","referenceUrls":["https://blogs.nvidia.com/"],"path":"/entities/nvidia","json":"/entities/nvidia.json","markdown":"/entities/nvidia.md"},{"id":"figure","name":"Figure","kind":"organization","context":"Figure develops humanoid robots and publishes its own tests of how they perform physical tasks.","referenceUrls":["https://www.figure.ai/news","https://www.figure.ai/news/helix-2-5-zero-shot-30-home-generalization"],"path":"/entities/figure","json":"/entities/figure.json","markdown":"/entities/figure.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"153cc8721a71d0d17f78645be14915dd77b0b082ad13f374bc98491a188ee7d7","revision":"c5432fb2-f7f2-4d3e-8577-011c3a71fcc4","revisionKind":"publication-id","title":"Hugging Face article presents Reson8’s Rumble-2-Flash speech model","url":"https://escapemeridian.com/readings/hugging-face-article-presents-reson8-s-rumble-2-flash-speech-153cc872"},"source":{"type":"source","id":"https://huggingface.co/blog/rumble-2-flash","revision":"534bf6bbac111d013823ea644911face72ac5c197e68c86057de56a40d213234","revisionKind":"sha256-source-text","title":"Rumble-2-Flash: an open 0.5B speech model for 40+ languages","url":"https://huggingface.co/blog/rumble-2-flash"},"status":"available","items":[],"next":null,"scope":"Explicit citations in public readings to this reading. Shared topics are not citations. Other record types need an explicit reference adapter."},"revisions":[{"id":"153cc8721a71d0d17f78645be14915dd77b0b082ad13f374bc98491a188ee7d7","title":"Hugging Face article presents Reson8’s Rumble-2-Flash speech model","argument":"Hugging Face’s article announces Reson8’s open speech recognition model Rumble-2-Flash, with 0.5B parameters and support for more than 40 languages. Reson8 says it competes with models 3–60x its size and runs in Transformers and vLLM without custom code. The team says it used on-policy distillation to make a smaller model: the student transcribes audio, then learns from a larger model’s token-by-token feedback. GOLD handles differences between the models’ tokenizers. On the Open ASR Leaderboard multilingual track, Reson8 reports a 6.96% average error rate across nine languages; on the English track, it reports 6.6% WER. On AppTek’s benchmark of 302 hours of helpline calls in 14 English accents, the model scored 7.86% WER, compared with 7.78% for NVIDIA’s Nemotron 3.5 Transcribe. Reson8 also reports throughput above 7,000 RTFx on larger leaderboard datasets with full batches, equivalent to about half a second for an hour of audio. The article presents these as benchmark results; it does not establish performance on other audio or in other deployment settings.","topicIds":["intelligence-and-discovery","work-and-agency"],"domain":"work","claimType":"attributed reading","limits":"The performance figures are claims reported in the post and are tied to the named benchmarks and conditions; they do not establish performance on other audio or in other deployment settings. The post reports throughput for larger leaderboard datasets where batches stay full, so that figure is not presented as a result for every workload. ","speaker":{"name":"Reson8, Mark 't Hart, Eustache Le Bihan, Eric Bezzam","roleAtSource":"Article author"},"source":{"title":"Rumble-2-Flash: an open 0.5B speech model for 40+ languages","publisher":"Hugging Face","format":"Article","url":"https://huggingface.co/blog/rumble-2-flash","publishedAt":"2026-11-18T00:00:00.772Z","dateBasis":"Date stated by the publisher."},"evidence":[{"text":"Reson8 announced Rumble-2-Flash as an open speech recognition model with 0.5B parameters and support for more than 40 languages.","excerpt":"Today we're open sourcing Rumble-2-Flash under the Apache 2.0 license. It's a 0.5B-parameter speech recognition model for 40+ languages","url":"https://huggingface.co/blog/rumble-2-flash","relationship":"Source-stated"},{"text":"Reson8 says the model was trained with on-policy distillation, in which the student transcribes audio and receives token-by-token feedback from the teacher.","excerpt":"The student doesn't train on transcripts the teacher wrote. It transcribes the audio itself, and the teacher grades that transcript token by token","url":"https://huggingface.co/blog/rumble-2-flash","relationship":"Source-stated"},{"text":"The post reports a 6.96% average error rate on the Open ASR Leaderboard multilingual track across nine named languages, and 6.6% WER on its English track.","excerpt":"Rumble-2-Flash averages a 6.96% error rate across Chinese, German, French, Italian, Spanish, Portuguese, Hindi, Dutch and English. That's better than Meta's omniASR-LLM 7B","url":"https://huggingface.co/blog/rumble-2-flash","relationship":"Source-stated"},{"text":"On AppTek’s call-centre benchmark, Reson8 reports 7.86% WER for Rumble-2-Flash and 7.78% for NVIDIA’s Nemotron 3.5 Transcribe.","excerpt":"Rumble-2-Flash reaches 7.86% WER, on par with NVIDIA's Nemotron 3.5 Transcribe (30B) at 7.78%.","url":"https://huggingface.co/blog/rumble-2-flash","relationship":"Source-stated"},{"text":"Reson8 reports throughput above 7,000 RTFx on larger leaderboard datasets when batches stay full.","excerpt":"on the larger datasets in HF's Open ASR Leaderboard, where batches stay full, it reaches over 7,000 RTFx.","url":"https://huggingface.co/blog/rumble-2-flash","relationship":"Source-stated"}],"sourceMedia":{"status":"found","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","checkedAt":"2026-11-18T19:50:11.127Z","items":[],"note":"Media was declared by the source, but no asset passed the supported availability checks. Open the original source to view it."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How does Rumble-2-Flash perform on audio and in deployment settings beyond the benchmarks described in the article?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Hugging Face article presents Reson8’s Rumble-2-Flash speech model","summary":"Hugging Face’s article announces Reson8’s open speech recognition model Rumble-2-Flash, with 0.5B parameters and support for more than 40 languages. Reson8 says it competes with models 3–60x its size and runs in Transformers and vLLM without custom code. The team says it used on-policy distillation to make a smaller model: the student transcribes audio, then learns from a larger model’s token-by-token feedback. GOLD handles differences between the models’ tokenizers. On the Open ASR Leaderboard multilingual track, Reson8 reports a 6.96% average error rate across nine languages; on the English track, it reports 6.6% WER. On AppTek’s benchmark of 302 hours of helpline calls in 14 English accents, the model scored 7.86% WER, compared with 7.78% for NVIDIA’s Nemotron 3.5 Transcribe. Reson8 also reports throughput above 7,000 RTFx on larger leaderboard datasets with full batches, equivalent to about half a second for an hour of audio. The article presents these as benchmark results; it does not establish performance on other audio or in other deployment settings.","question":"How does Rumble-2-Flash perform on audio and in deployment settings beyond the benchmarks described in the article?","claims":[{"text":"Reson8 announced Rumble-2-Flash as an open speech recognition model with 0.5B parameters and support for more than 40 languages.","excerpt":"Today we're open sourcing Rumble-2-Flash under the Apache 2.0 license. It's a 0.5B-parameter speech recognition model for 40+ languages","support":"source-stated"},{"text":"Reson8 says the model was trained with on-policy distillation, in which the student transcribes audio and receives token-by-token feedback from the teacher.","excerpt":"The student doesn't train on transcripts the teacher wrote. It transcribes the audio itself, and the teacher grades that transcript token by token","support":"source-stated"},{"text":"The post reports a 6.96% average error rate on the Open ASR Leaderboard multilingual track across nine named languages, and 6.6% WER on its English track.","excerpt":"Rumble-2-Flash averages a 6.96% error rate across Chinese, German, French, Italian, Spanish, Portuguese, Hindi, Dutch and English. That's better than Meta's omniASR-LLM 7B","support":"source-stated"},{"text":"On AppTek’s call-centre benchmark, Reson8 reports 7.86% WER for Rumble-2-Flash and 7.78% for NVIDIA’s Nemotron 3.5 Transcribe.","excerpt":"Rumble-2-Flash reaches 7.86% WER, on par with NVIDIA's Nemotron 3.5 Transcribe (30B) at 7.78%.","support":"source-stated"},{"text":"Reson8 reports throughput above 7,000 RTFx on larger leaderboard datasets when batches stay full.","excerpt":"on the larger datasets in HF's Open ASR Leaderboard, where batches stay full, it reaches over 7,000 RTFx.","support":"source-stated"}],"limitations":["The performance figures are claims reported in the post and are tied to the named benchmarks and conditions; they do not establish performance on other audio or in other deployment settings.","The post reports throughput for larger leaderboard datasets where batches stay full, so that figure is not presented as a result for every workload. "],"mediaId":"5b89a2db7f60ae9099b72ca73768977f61c082125cbaf94b72c9c9bced5f17aa","correctionReason":null,"id":"153cc8721a71d0d17f78645be14915dd77b0b082ad13f374bc98491a188ee7d7","publicationId":"c5432fb2-f7f2-4d3e-8577-011c3a71fcc4","version":1,"sourceUrl":"https://huggingface.co/blog/rumble-2-flash","sourceTitle":"Rumble-2-Flash: an open 0.5B speech model for 40+ languages","publisher":"Hugging Face","publishedAt":"2026-11-18T00:00:00.772Z","originalPublished":{"value":"2026-11-18T00:00:00.772Z","precision":"instant","basis":"publisher_page"},"firstObservedAt":"2026-11-18T14:43:04.732Z","createdAt":"2026-11-18T19:34:19.294Z","historical":false,"entityIds":["hugging-face"],"topicIds":["intelligence-and-discovery","work-and-agency"],"media":[{"id":"5b89a2db7f60ae9099b72ca73768977f61c082125cbaf94b72c9c9bced5f17aa","kind":"article","url":"https://huggingface.co/blog/rumble-2-flash","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-18T19:33:08.199Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"found","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","checkedAt":"2026-11-18T19:50:11.127Z","items":[],"note":"Media was declared by the source, but no asset passed the supported availability checks. Open the original source to view it."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v9","mediaId":"5b89a2db7f60ae9099b72ca73768977f61c082125cbaf94b72c9c9bced5f17aa","confidence":0.53,"provider":"typesafe","model":"jev-1.13.0","inputHash":"09dc399502e8d1666f7e258f14a019dfc71a19db846889a7a7c5c93c4fa0d10c","decidedAt":"2026-11-18T19:33:11.257Z"},"provenance":{"textHash":"534bf6bbac111d013823ea644911face72ac5c197e68c86057de56a40d213234","observationId":"9af96bcd1e8c2caa39933ac19f0102ef7576081e66b20658eb6de013b827ff6c","retrievedAt":"2026-11-18T19:33:07.792Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"553f8618-397a-4779-8663-a26a18427f1e"},"reviewer":{"model":"gpt-6-luna","runId":"1474d691-c893-4fe5-89e3-6358db91d3cf"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-18T19:34:19.294Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Reson8","url":null,"kind":"Person"},{"name":"Mark 't Hart","url":null,"kind":"Person"},{"name":"Eustache Le Bihan","url":"https://huggingface.co/eustlb","kind":"Person"},{"name":"Eric Bezzam","url":"https://huggingface.co/bezzam","kind":"Person"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Reson8","url":null,"kind":"Person","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Reson8","status":"stated","provenance":{"basis":"article-jsonld","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","locator":"document/html[1]/head[0]/script[43]@3226.author"},"assertionId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-18T19:50:11.127Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","contentSha256":"2afea669ef2d2382c1c089cabf7466f418cb86c370f51a872e53b09bac276675","textHash":"534bf6bbac111d013823ea644911face72ac5c197e68c86057de56a40d213234","observedAt":"2026-11-18T19:50:11.127Z"}},{"actor":{"name":"Mark 't Hart","url":null,"kind":"Person","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Mark 't Hart","status":"stated","provenance":{"basis":"article-jsonld","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","locator":"document/html[1]/head[0]/script[43]@3226.author"},"assertionId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-18T19:50:11.127Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","contentSha256":"2afea669ef2d2382c1c089cabf7466f418cb86c370f51a872e53b09bac276675","textHash":"534bf6bbac111d013823ea644911face72ac5c197e68c86057de56a40d213234","observedAt":"2026-11-18T19:50:11.127Z"}},{"actor":{"name":"Eustache Le Bihan","url":"https://huggingface.co/eustlb","kind":"Person","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Eustache Le Bihan","status":"stated","provenance":{"basis":"article-jsonld","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","locator":"document/html[1]/head[0]/script[43]@3226.author"},"assertionId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-18T19:50:11.127Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","contentSha256":"2afea669ef2d2382c1c089cabf7466f418cb86c370f51a872e53b09bac276675","textHash":"534bf6bbac111d013823ea644911face72ac5c197e68c86057de56a40d213234","observedAt":"2026-11-18T19:50:11.127Z"}},{"actor":{"name":"Eric Bezzam","url":"https://huggingface.co/bezzam","kind":"Person","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Eric Bezzam","status":"stated","provenance":{"basis":"article-jsonld","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","locator":"document/html[1]/head[0]/script[43]@3226.author"},"assertionId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-18T19:50:11.127Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","contentSha256":"2afea669ef2d2382c1c089cabf7466f418cb86c370f51a872e53b09bac276675","textHash":"534bf6bbac111d013823ea644911face72ac5c197e68c86057de56a40d213234","observedAt":"2026-11-18T19:50:11.127Z"}},{"actor":{"name":"Hugging Face","url":"https://huggingface.co/","kind":"unknown","actorId":null},"role":"publisher","scope":{"kind":"publication","locator":null},"literalCredit":"Hugging Face","status":"stated","provenance":{"basis":"article-publisher","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","locator":"document/html[1]/head[0]/script[43]@3226.publisher"},"assertionId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-18T19:50:11.127Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00","sourceUrl":"https://huggingface.co/blog/rumble-2-flash","contentSha256":"2afea669ef2d2382c1c089cabf7466f418cb86c370f51a872e53b09bac276675","textHash":"534bf6bbac111d013823ea644911face72ac5c197e68c86057de56a40d213234","observedAt":"2026-11-18T19:50:11.127Z"}}],"participationCoverage":{"total":5,"returned":5,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":{"name":"Hugging Face","url":"https://huggingface.co/"},"sourceUrl":"https://huggingface.co/blog/rumble-2-flash","snapshotId":"da14ade2a14c95e356d09ee8387c54de0ed4600731587e64d4433534bdeaca00","contentSha256":"2afea669ef2d2382c1c089cabf7466f418cb86c370f51a872e53b09bac276675","textHash":"534bf6bbac111d013823ea644911face72ac5c197e68c86057de56a40d213234","observedAt":"2026-11-18T19:50:11.127Z","basis":"publisher-page-metadata","sourceLinks":50,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/153cc8721a71d0d17f78645be14915dd77b0b082ad13f374bc98491a188ee7d7/source"}}}],"audits":[{"publicationId":"c5432fb2-f7f2-4d3e-8577-011c3a71fcc4","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The post is a Hugging Face blog article co-authored by Reson8 and Hugging Face contributors. The 'we' announcing the release is the Reson8 team, and the reading attributes performance claims to Reson8.","excerpt":"Today we're open sourcing Rumble-2-Flash under the Apache 2.0 license."},{"dimension":"summary","status":"uncertain","explanation":"The facts are accurate, but a general reader is left with unexplained central jargon. WER is never defined as word error rate (the share of words transcribed wrongly). GOLD, tokenizer, vLLM and Transformers are not explained, and RTFx is only partly explained. The reading also drops the size contrast that gives the call-centre result its force: Nemotron is a 30B model and Rumble-2-Flash is 0.5B. Why the model might matter (small, open, fast) is never stated. Smallest repair: define WER and tokenizer in plain words, note Nemotron's 30B size, and add one source-supported sentence on significance.","excerpt":"on par with NVIDIA's Nemotron 3.5 Transcribe (30B) at 7.78%"},{"dimension":"date","status":"pass","explanation":"The originalPublished date, 2026-11-18, matches the date on the publisher page.","excerpt":"Published\nNovember 18, 2026"},{"dimension":"support","status":"pass","explanation":"All claims and figures match the source: 0.5B parameters, 40+ languages, 6.96% and 6.6% error rates, 7.86% vs 7.78% on the call-centre benchmark, and over 7,000 RTFx with full batches. The limitations and the question are appropriately conditional and do not invent benefits or costs.","excerpt":"Its throughput depends on how full the batches are"},{"dimension":"media","status":"pass","explanation":"The selected media is the verified link to the original article, which is the direct source.","excerpt":null}],"createdAt":"2026-11-18T19:34:19.294Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}