← Back to live feed1 story
DeepSeek has released a 763 billion parameter open-weight model under an MIT license, designed for agentic coding and task automation with a 1 million token context window. The model scored 77.0 on the DeepSWE benchmark, surpassing Claude Opus 5 and GPT-5.6 Sol, and is available across cloud providers including Baseten, Ollama, and OpenRouter. Off-peak pricing is $0.15 per million input tokens and $0.60 per million output tokens, significantly lower than Anthropic's rates for Opus 5.
Sign in to suggest
Key sources
- SOURCE@deepseek_ai“a Pareto improvement in performance and cost. Now live on our website, app, and API.”x.com
- SOURCE@deepseek_ai“Get DeepSeek-V4.1-Flash-1017: 🤗 Hugging Face: https://t.co/xA4zPuxLtq 🔗 ModelScope: https://t.co/Zsg7FKld6Q”x.com
- SUPPORT@thom_wolf“moving Terminal-Bench 4.0 from 31.2 to 42.3”x.com
- SUPPORT@kimmonismus“beats GPT‑5.6 Sol and Opus 5 on selected coding and agent benchmarks, at much lower token prices”x.com
- SUPPORT@pawelhuryn“763B params. The weights are already on Hugging Face.”x.com
- SUPPORT@teortaxestex“Notice the cost column”x.com
- SUPPORT@jun_song“The V4 Flash release was a much bigger shock, and the market barely moved.”x.com
- SUPPORT@martinvars“how anyone will ever make the $8 trillion in AI infrastructure back.”x.com