# OpenAI Prices GPT-6.1 Sol at One Fifth of Astra, Matching Anthropic's Sonnet Rate

The model is priced at $2 per million input tokens and $10 per million output tokens, with a 1.05 million token context window, the same headline rate Anthropic set for Claude Sonnet 5.5 one day earlier.

- Published: 2026-09-30T06:34:52.927Z
- Canonical: https://polylog.news/ai/2026-09-30/openai-prices-gpt-6-1-sol-at-one-fifth-of-astra-matching-ant
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [OpenAI News](https://openai.com/index/introducing-gpt-6-1-sol), [OpenAI DevDay 2026 Recap](https://openai.com/index/devday-2026-recap), [Polylog editors](https://polylog.news)

OpenAI used its DevDay keynote on Monday to reprice the middle of its model lineup. [GPT-6.1 Sol](https://openai.com/index/introducing-gpt-6-1-sol) is positioned as offering near-Astra intelligence for agentic coding, computer use and professional work, at one-fifth of GPT-6 Astra's standard input and output token prices. The published rates are [$2 per million input tokens, $0.10 per million cached input tokens and $10 per million output tokens](https://www.datacamp.com/blog/gpt-6-1-sol), with a 1,050,000-token context window and a maximum output of 128,000 tokens. The cached-input rate is half what GPT-6 Sol charged a week earlier.

The more important comparison is external. Anthropic released Claude Sonnet 5.5 the day before at the same $2 and $10 per million tokens, so the two most heavily used mid-tier coding models now carry identical list prices. VentureBeat also reports [an ultrafast serving tier running at about 300 tokens per second](https://venturebeat.com/technology/openais-gpt-6-1-sol-offers-astra-like-performance-at-1-5th-price-a-new-ultrafast-tier-clocks-at-300-tokens-per-second), which is likely where competition will shift once the two vendors' prices converge.

"Near-Astra" is OpenAI's own characterization of the model. The company has not published a third-party replication, and independent reference points, such as agentic-coding and computer-use test suites run by outside evaluators, will take days to appear. The model is available now for Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex, and in the API as `gpt-6.1-sol`, alongside [more than twenty other DevDay announcements](https://openai.com/index/devday-2026-recap). OpenAI told the audience that ChatGPT now reaches [1.2 billion weekly users](https://t.me/aipost/8309), a distribution base large enough that a list-price cut directly reduces gross margin rather than functioning mainly as a marketing move.

## What this means

The mechanism is substitution among models of similar capability. When a vendor sells most of the frontier's usefulness at one-fifth of the price, the first buyers to move are high-volume agentic workloads, where token volume rather than peak reasoning ability determines the bill. Exposed are Anthropic's Sonnet tier, which now competes at the same list price without OpenAI's consumer distribution advantage, and every inference provider whose margin assumed a stable price floor. The precedent is the 2024-2026 pattern in which each cheap-tier launch was answered within weeks by a rival's repricing rather than by a new capability. The likelier outcome is another below-rival price cut or cache-discount move from a major API vendor within weeks, with Google and Meta the most probable movers given their own serving-cost advantages. This would be proved wrong if independent agentic-coding evaluations show Sol performing materially below Astra on real tasks, in which case the price cut would represent a downgrade in tier rather than a price war.

## What to watch

- Independent agentic-coding and computer-use evaluations of GPT-6.1 Sol against GPT-6 Astra and Claude Sonnet 5.5, which will show whether the price cut came with a reduction in capability.
- Whether Anthropic or Google responds with its own repricing rather than a new model. That would be the clearest sign that competition has shifted from capability to unit cost.
- Cached-input pricing across vendors. Cache discounts now move faster than headline rates and determine the economics of long-running agents.