September 29, 2026
GPT-6.1 Sol nearly caught Astra: $0.72 per Intelligence Index task
On Sep 29, GPT-6.1 Sol trailed GPT-6 Astra by 1 point in the Intelligence Index, but an evaluated task cost $0.72 instead of $3.26. In the API, `gpt-6.1-sol` costs $2 per 1 million input tokens, $0.10 for cached input, and $10 for output. In ChatGPT Work and Codex, the model is available on Plus, Pro, Business, Enterprise, and Edu plans.
Artificial Analysis
@artificialanlys
GPT-6.1 Sol replaced GPT-6 Sol just 7 days later. In the Intelligence Index, it trails GPT-6 Astra by 1 point, while a task costs less than a quarter of the price. Pricing is the same as GPT-6 Sol: $2/$10 per million input/output tokens. There is one exception: the discount for cache reads rose from 90% to 95%. As a result, GPT-6.1 Sol's blended cost in agentic tasks is slightly lower than GPT-6 Sol's. This is another price reduction after GPT-6 Sol's initial 50% discount relative to GPT-5.6 Sol. Key points: ➤ Nearly Astra-level: GPT-6.1 Sol added 4 points in the Intelligence Index relative to GPT-6 Sol and 5 points relative to GPT-5.6 Sol. It stopped 1 point short of GPT-6 Astra. In knowledge-work agents, the model gained 4 points in AA-Briefcase v1.1 and 5 points in GDPval-AA v2.1. Other results include: +12 points in Terminal-Bench 4.0, +5 in Humanity’s Last Exam, +6 in GDP.pdf, and +8 in AA-Omniscience Accuracy. The hallucination rate fell from 60% to 54%. ➤ A new price frontier: at maximum reasoning effort, GPT-6.1 Sol spends less than a quarter of GPT-6 Astra's cost per Intelligence Index task: $0.72 versus $3.26. That is 31% cheaper than GPT-6 Sol at $1.05 per task and 64% cheaper than GPT-5.6 Sol at $1.99. At every reasoning effort level, GPT-6.1 Sol extends the price Pareto frontier: no model is cheaper at the same intelligence level. ➤ A new token frontier, though output tokens are slightly higher: at different reasoning effort levels, GPT-6.1 Sol uses 10–30% more output tokens than GPT-6 Sol. But thanks to its higher Intelligence Index score, its low and medium levels have become optimal for the balance of tokens and results. ➤ Gains in the Coding Agent Index: at maximum reasoning effort, GPT-6.1 Sol added 3 points over GPT-6 Sol in the Artificial Analysis Coding Agent Index and trailed GPT-6 Astra by 2 points. Congratulations to OpenAI and sama on the launch!
· 16.1K views
Earlier, GPT-6 Sol already cost 50% less than GPT-5.6 Sol. Now GPT-6.1 Sol added 4 points in the Intelligence Index, while Artificial Analysis estimates $1.05 per task versus $1.99 for GPT-5.6 Sol.
Agent pricing. Input and output cost the same as GPT-6 Sol, but cached input is half the price: $0.10 instead of $0.20 per 1 million tokens. Artificial Analysis says the blended cost of agentic tasks is now lower, even though the model uses 10–30% more output tokens.
GitHub Copilot has already begun a gradual rollout of GPT-6.1 Sol for Pro+, Max, Business, and Enterprise.
