--- format: "aidr-story-markdown/v1" id: "36d449366e99fab13a01be49c28c05aad309073f94e5325bac5c088547ec2461" canonical_url: "https://aidr.today/36d44936?lang=en" title: "Ornith Launches 397B Self Improving Model Family to Top Open Coding Benchmarks" lang: "en" requested_lang: "en" available_langs: ["en","vi"] translation_fallback: null fallback_fields: [] published_at: "2026-08-20T05:48:40.000Z" category: "Releases" topics: ["open-source","coding","agent","llm","inference"] source_urls: ["https://huggingnews.com/ai/ornith-launches-397b-self-improving-model-family-to-top-open-coding-benc-fc355347","https://x.com/vllm_project/status/2090243605147586955","https://x.com/ornith_/status/2090248152024801480","https://x.com/ornith_/status/2090276420983587087","https://x.com/sgl_project/status/2090292384408207654","https://x.com/ornith_/status/2090297133069107493","https://x.com/ornith_/status/2090258105682854390","https://x.com/ornith_/status/2090146869003669948"] summary: "Large language models featuring weights from 9B to 397B parameters are now available under an MIT license from the Ornith team. The Ornith-1.5 models use an end-to-end self-improvement loop in which the system generates its own tasks and solution rollouts for reinforcement learning. This approach allows the models to reach state of the art performance among open models for coding and agentic tasks. Inference platforms vLLM, Ollama, and SGLang integrated serving support for the models immediately following the launch. The vLLM and SGLang stacks currently support the 9B variant, while the full series is accessible via Ollama. Quantized versions in FP8, GGUF, MLX, and NVFP4 formats were released to facilitate deployment across diverse hardware." --- # Ornith Launches 397B Self Improving Model Family to Top Open Coding Benchmarks > [Open the canonical story]() **Published:** 2026-08-20T05:48:40.000Z **Category:** Releases **Topics:** open\-source, coding, agent, llm, inference ## Summary Large language models featuring weights from 9B to 397B parameters are now available under an MIT license from the Ornith team\. The Ornith\-1\.5 models use an end\-to\-end self\-improvement loop in which the system generates its own tasks and solution rollouts for reinforcement learning\. This approach allows the models to reach state of the art performance among open models for coding and agentic tasks\. Inference platforms vLLM, Ollama, and SGLang integrated serving support for the models immediately following the launch\. The vLLM and SGLang stacks currently support the 9B variant, while the full series is accessible via Ollama\. Quantized versions in FP8, GGUF, MLX, and NVFP4 formats were released to facilitate deployment across diverse hardware\. ## Sources - [Story source]() - [Story source]() - [Supporting source]() - [Story source]() - [Supporting source]() - [Story source]() - [Story source]() - [Supporting source]()