Breaking

OpenAI prices GPT-6.1 Sol at one-fifth of Astra for near-flagship work

September 29, 2026 at 15:42 EDT

Foundation Models
AI Agents
Software Dev & Coding

OpenAI released GPT-6.1 Sol on September 29, 2026, at its DevDay event, setting standard input and output prices at one-fifth of GPT-6 Astra. The company says the update nearly matches Astra on agentic coding, computer use, and professional work, and it prices cached input at $0.10 per million tokens — 95 percent below its own standard input rate.

September 29, 2026 · OpenAI

Near-Astra work at one-fifth the price

GPT-6.1 Sol keeps the $2 / $10 token rates of GPT-6 Sol, halves cached input, and is aimed at agentic coding, computer use, and professional work that previously defaulted to Astra.

Standard price per 1M tokens

Input price, drawn to scale

Each block is $2. Output follows the same five-to-one gap: $50 for Astra, $10 for GPT-6.1 Sol.

$10

GPT-6 Astra input

GPT-6.1 Sol input

5× the input price for the flagship, for a model OpenAI says matches it on a hard coding benchmark.

Cached input per 1M tokens

Reused context, block by block

Each block is $0.10. Sol's new cache rate is half of GPT-6 Sol and one-tenth of Astra.

$1.00

GPT-6 Astra

$0.20

GPT-6 Sol

$0.10

GPT-6.1 Sol

That $0.10 rate is 95% below Sol's own $2 standard input price.

5×

Astra vs Sol on standard input and output

+6.4 pts

DeepSWE versus GPT-6 Sol's best score

2.1 pts

OSWorld gap still left to Astra at max effort

68.1%

Astra's lead score on Terminal-Bench Science

Why teams switch

Same $2 / $10 rates as GPT-6 Sol, with higher scores on the published charts. Cached context at $0.10 helps agents that resend a long prompt. Early Copilot tests used fewer tokens and steps.

What still holds it back

Astra still leads the hardest science and the top computer-use score. Regular Chat does not have it. Tool calls must leave Chat Completions, and a prompt over 272K tokens reprices the whole request.

Where it runs

Work and Codex today. Chat is still closed.

Paid plans

Plus, Pro, Business, Enterprise, and Edu in ChatGPT Work and Codex

API id

gpt-6.1-sol, text and image in, text out

Not yet

Ordinary Chat. Ultrafast, up to 8× in Codex, is due in the coming days.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.