Cohere released three open-source AI models in 2025 under Apache 2.0. The models include Transcribe, Command A+, and North Mini Code, targeting speech, language, and code tasks.
Key facts
- Three models released: Transcribe, Command A+, North Mini Code
- All under Apache 2.0 license
- Models target speech, language, and code tasks
- Cohere did not disclose model sizes or benchmarks
- Release year: 2025
Cohere has released open-source models Transcribe, Command A+, and North Mini Code so far this year, all available under the Apache 2.0 license According to @_akhaliq. The announcement, made via a retweet from the Cohere account, marks a significant expansion of the company's open-source portfolio.
Transcribe is an automatic speech recognition model, Command A+ targets general language understanding and generation, and North Mini Code is optimized for code generation tasks. By releasing these under Apache 2.0, Cohere permits commercial use, modification, and redistribution with minimal restrictions—a more permissive stance than many competitors who use custom licenses or restrict commercial use.
This open-source push comes as the enterprise AI market increasingly demands transparency and customization. Cohere, which has long positioned itself as the enterprise-friendly alternative to OpenAI and Anthropic, is betting that Apache 2.0 licensing will accelerate adoption among companies wary of vendor lock-in or restrictive terms.
The company did not disclose specific model sizes, benchmark scores, or training costs for any of the three models. This lack of technical detail makes it difficult to compare performance against alternatives like Whisper (speech), Llama 3 (language), or Code Llama (code).
Cohere's strategy mirrors a broader industry trend toward open-weight models, popularized by Meta's Llama series and Mistral AI. However, Cohere differentiates by targeting enterprise workflows directly, with models designed for integration into business applications rather than general-purpose research.
The timing is notable: Cohere has been relatively quiet on the open-source front compared to rivals. This release could help the company regain momentum in the developer community, which has increasingly gravitated toward open-weight models from Meta, Mistral, and others.
What to watch
Watch for Cohere to release benchmark scores or technical papers for these models. If performance is competitive, enterprise adoption could accelerate. Also watch for pricing tiers or hosted API versions that monetize the open-source base.
