OpenAI releases GPT-6.1 Sol with near-Astra intelligence at 80% off

OpenAI just dropped a model that delivers near-frontier Astra performance at a fifth of the price.

OpenAI announced GPT-6.1 Sol, an upgraded mid-tier model built for agentic coding, computer use, and business workflows. According to the release, Sol nearly matches flagship GPT-6 Astra across key benchmarks while cutting input and output token prices by 80%. Standard API access runs $2 per million input tokens and $10 per million output, with cached inputs dropping to ten cents.

Why it matters: Here's the gist: autonomous agents get expensive fast when chewing through large codebases and complex documents. By dropping cached input costs to $0.10 per million tokens and matching top-tier models on coding benchmarks, running multi-step background agents becomes significantly cheaper to build and operate.

Know this:

• Availability: Live today via the API as gpt-6.1-sol, plus ChatGPT Work and Codex. Standard Chat access is not available yet.
• Benchmarks: On the DeepSWE v1.1 coding benchmark, Sol matches GPT-6 Astra at roughly one-fifth the cost. On Terminal-Bench Science, average task costs fell from $23.80 on Astra to $5.47 on Sol.
• Speed bump: OpenAI plans to release an "Ultrafast" version in the coming days, promising up to 8x faster token generation in Codex.

Your monthly API bill might finally stop looking like a phone number.