← Back to live feed1 story
OpenAI has integrated the GPT-6.1 Terra model into GitHub Copilot's CLI and app, making it available to developers via OpenRouter. The model, designed for visual reasoning and computer use, solved 50.3% of tasks on the Terminal-Bench 4.0 coding benchmark and reduced hallucination rates to 38% on the AA-Omniscience benchmark. It is priced at $1 per million input tokens and $5 per million output tokens, with a 1.05M context window, and OpenAI has reclassified ChatGPT Work's default power presets to use GPT-6.1 Terra at the Medium setting.
Sign in to suggest
Key sources
- SOURCE@openai“finishes tasks over 3× faster on average than both Astra and GPT-6.1 Sol”x.com
- SOURCE@openrouter“$1/M input, $5/M output, 1.05M context.”x.com
- SOURCE@digg“available now in ChatGPT Work and Codex and through the API.”x.com
- SUPPORT@artificialanlys“GPT-6.1 Terra guesses 38% of the time when it does not know the answer to a question in AA-Omniscience... Previously, GPT-5.6 Terra's hallucination rate was 88%”x.com
- SUPPORT@teortaxestex“The least hallucinatory OpenAI model?”x.com
- SUPPORT@mkeremturhan“Cut $200 ChatGPT Pro Codex allowance from 20x to 10x Plus this Friday (Oct 30).”x.com
- SUPPORT@davis7“I don’t rly think this release for us... unless ur building an agent into a product”x.com
- SUPPORT@designarena“GPT-6.1 Terra took almost 4x longer and used 48% more tokens, but it still came out 38% cheaper”x.com