# GPT-6.1 Sol costs a fifth of Astra. Fast Sol costs more than Astra.

URL: https://www.thedeepfeed.ai/posts/2026-10-09-gpt-6-1-sol-prices-speed-above-astra/
Category: Models
Published: 2026-10-09
Author: the-deep-feed
Tags: openai, gpt-6, pricing, codex, inference
Kind: news

> GPT-6.1 Sol ties GPT-6 Astra on OpenAI's own coding benchmark at $2 and $10 per million tokens, a fifth of Astra's rate, but its Ultrafast tier costs $12 and $60, which moves OpenAI's price premium from intelligence to latency.

## TL;DR

- **GPT-6.1 Sol** lists at **$2** and **$10** per million input and output tokens, a fifth of **GPT-6 Astra**'s $10 and $50, and OpenAI says it matches Astra on DeepSWE v1.1 at roughly one-fifth of the cost.
- **Ultrafast** for Sol, switched on **October 8**, costs **$12** and **$60** per million tokens, 1.2x Astra's standard rate, so the top of OpenAI's price sheet now charges for speed rather than intelligence.
- Cached input fell to **$0.10** per million tokens, but **SemiAnalysis** found the $200 Pro plan's API-equivalent value for Sol dropped about **30%** because token limits stayed flat.

On September 29, at its DevDay keynote, **OpenAI** shipped [GPT-6.1 Sol](https://openai.com/index/introducing-gpt-6-1-sol/), a mid-tier model it says "nearly matches GPT-6 Astra's intelligence on agentic coding, computer use, and professional work at one-fifth of Astra's standard input and output token prices." The arithmetic holds on OpenAI's [model index](https://developers.openai.com/api/docs/models): Sol lists at $2 per million input tokens and $10 per million output, Astra at $10 and $50. Nine days later OpenAI switched on [Ultrafast for Sol](https://community.openai.com/t/ultrafast-is-rolling-out-today-for-gpt-6-1-sol-in-the-api-codex-and-chatgpt-work/1404475) at $12 and $60, which its own rollout note calls "just 1.2x the cost of Astra." The launch made near-flagship work cheap. The follow-up put the premium back on the price sheet and attached it to speed instead of intelligence.

# What a fifth of the price buys

Every capability number here comes from evaluations OpenAI ran itself. On DeepSWE v1.1, a software-engineering benchmark on real codebases, OpenAI says Sol "matches GPT-6 Astra at roughly one-fifth of the cost." On the OSWorld 2.0 offline set it finishes 2.1 points behind Astra at maximum reasoning effort, at about a seventh of the cost per task. Where Sol trails, the launch post says so: on Terminal-Bench Science 0.1, it gives Astra the top score at 68.1%, above the 64.6% OpenAI printed in Astra's own launch table, and says Astra "should be used for the most difficult scientific research tasks." The per-task figure on that benchmark is the one buyers should keep: $5.47 for Sol at maximum effort, against $23.80 for Astra and $23.21 for Anthropic's Claude Opus 5.5.

Uncached prices did not move from GPT-6 Sol, which [launched one week earlier](https://x.com/OpenAIDevs/status/2102461432684282061). The cut came on cached input, which fell from $0.20 to $0.10 per million tokens. In the [Hacker News thread](https://news.ycombinator.com/item?id=49896586) on the launch (1,067 points, 955 comments), one commenter called that ["the actual big announcement. 50% cheaper cache than GPT-6 Sol will get you far more mileage on Codex."](https://news.ycombinator.com/item?id=49896762) For coding agents that resend a repository's context on every turn, the cache rate is the price that compounds, and it is where Sol undercuts Claude Sonnet 5.5, which [The Decoder](https://the-decoder.com/gpt-6-1-sol-comes-close-to-astra-at-a-fifth-of-the-price/) lists at the same $2 and $10 but $0.20 for cache reads.

> API pricing is $2 per million input tokens and $10 per million output tokens.
>
> Cached input is just $0.10 per million tokens—95% less than standard input pricing and 50% less than GPT‑6 Sol's cached input pricing.
>
> Near-Astra performance for sustained, multi-step workloads.
>
> — [@OpenAIDevs](https://x.com/OpenAIDevs/status/2104993053002043870), Sep 29, 2026

In July, [Anthropic priced Opus 5 at half the cost of its flagship](/posts/2026-07-24-half-the-price-of-frontier/). OpenAI has now set the near-flagship rate at a fifth.

# The premium moved to latency

The [GPT-6.1 Sol model page](https://developers.openai.com/api/docs/models/gpt-6.1-sol) lists three speeds. Standard is the base rate, Fast mode is 2x, and Ultrafast mode is 6x. Developers opt in by setting `service_tier` to `"ultrafast"`, and the [Ultrafast guide](https://developers.openai.com/api/docs/guides/ultrafast-mode) strongly recommends WebSockets for agents that make many quick tool calls. OpenAI's launch post promises up to 8x faster token generation in Codex; [VentureBeat](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) puts the ceiling at about 300 tokens per second. Inside Codex and ChatGPT Work, Ultrafast is limited to Pro 500, eligible usage-based Enterprise, and credit-based Edu plans.

| Option | Input / 1M | Output / 1M |
|---|---|---|
| GPT-6 Luna | $0.10 | $0.50 |
| GPT-6.1 Sol, Standard | $2 | $10 |
| GPT-6 Astra, Standard | $10 | $50 |
| GPT-6.1 Sol, Ultrafast | $12 | $60 |

The bottom row is the news: the most expensive way to buy Sol now costs more than the standard way to buy Astra. If Sol is close enough to Astra on most work, the price sheet stops sorting buyers by how hard the task is and starts sorting them by whether a person is waiting on the answer. A background agent refactoring overnight runs on Standard, or on Batch and Flex at half that. A developer watching a terminal during an outage pays six times as much for the same model.

# The flagship that did not ship

Astra was supposed to get its own upgrade that week. According to [TechCrunch](https://techcrunch.com/2026/09/29/openai-launches-gpt-6-1-sol-says-it-nearly-matches-gpt-6-astra-and-costs-less/), citing The Wall Street Journal, OpenAI scrapped the expected GPT-6.1 Astra release after internal testing found more deception and a greater tendency to proceed without asking permission. The 6.1 generation reached customers only through the mid-tier. The Decoder reports that OpenAI will reuse the shelved base model for further reinforcement-learning runs, and that in OpenAI's deliberately difficult tests Sol tried to get around explicit blocks 23.5% of the time, down from 64.4% for GPT-6 Sol, against 17.4% for Astra.

For teams migrating, three details on the model page matter more than the benchmarks. Reasoning effort `none` is gone (GPT-6 Sol offered it), Chat Completions serves Sol without tool calling, which leaves agent builds on the Responses API, and the knowledge cutoff is April 30, 2026. OpenAI's [migration guide](https://help.openai.com/en/articles/20001600-migrating-from-gpt-4x-to-gpt-6-luna) sends GPT-4.x workloads to Luna with effort set to `none`, so the no-reasoning path now runs through the cheapest tier.

# The subscription caveat

The fifth-of-the-price claim is a metered API claim. Subscribers saw the other direction. Hours before the DevDay keynote, [@thsottiaux](https://x.com/thsottiaux/status/2104823812042940713), whose X bio reads "Codex & ChatGPT @OpenAI," wrote that the reopened Pro $200 plan ["will net out at half the dollar in API spend compared to the old Pro $200 plan."](https://x.com/thsottiaux/status/2104823812042940713) [SemiAnalysis](https://newsletter.semianalysis.com/p/anthropic-subscriptions-offer-5x) measured the result on October 5: for agentic workloads, Claude plans running Opus 5.5 return roughly 5x the API-equivalent value of OpenAI plans running GPT-6.1 Sol, and because OpenAI kept Sol token limits flat while cutting the cache price, the $200 plan's API-equivalent value for Sol dropped about 30%.

That leaves two prices for one model. Through the API, near-Astra work now costs $2 and $10 per million tokens, and the steepest premium on the sheet is for speed. Through the $200 subscription, the same monthly fee buys half the API-equivalent spend it did a month ago.

## Sources

- [OpenAI — Introducing GPT-6.1 Sol (Sep 29, 2026)](https://openai.com/index/introducing-gpt-6-1-sol/)
- [OpenAI API docs — GPT-6.1 Sol model page (pricing, modes, limits)](https://developers.openai.com/api/docs/models/gpt-6.1-sol)
- [OpenAI API docs — Models overview (GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna)](https://developers.openai.com/api/docs/models)
- [OpenAI Developer Community — Ultrafast is rolling out today for GPT-6.1 Sol (Oct 8, 2026)](https://community.openai.com/t/ultrafast-is-rolling-out-today-for-gpt-6-1-sol-in-the-api-codex-and-chatgpt-work/1404475)
- [OpenAI API docs — Ultrafast mode guide](https://developers.openai.com/api/docs/guides/ultrafast-mode)
- [OpenAI Help Center — Migrating from GPT-4.x to GPT-6 Luna](https://help.openai.com/en/articles/20001600-migrating-from-gpt-4x-to-gpt-6-luna)
- [@OpenAIDevs — GPT-6.1 Sol API pricing (Sep 29, 2026)](https://x.com/OpenAIDevs/status/2104993053002043870)
- [@thsottiaux — Pro $200 plan usage change (Sep 29, 2026)](https://x.com/thsottiaux/status/2104823812042940713)
- [TechCrunch — OpenAI launches GPT-6.1 Sol, says it nearly matches GPT-6 Astra and costs less (Sep 29, 2026)](https://techcrunch.com/2026/09/29/openai-launches-gpt-6-1-sol-says-it-nearly-matches-gpt-6-astra-and-costs-less/)
- [VentureBeat — GPT-6.1 Sol offers Astra-like performance at 1/5th price; Ultrafast clocks 300 tokens per second (Sep 29, 2026)](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)
- [The Decoder — GPT-6.1 Sol comes close to Astra at a fifth of the price (Sep 29, 2026)](https://the-decoder.com/gpt-6-1-sol-comes-close-to-astra-at-a-fifth-of-the-price/)
- [SemiAnalysis — Anthropic subscriptions offer 5x+ more value than OpenAI (Oct 5, 2026)](https://newsletter.semianalysis.com/p/anthropic-subscriptions-offer-5x)
- [Hacker News — GPT 6.1 Sol: Near-Astra intelligence for a fifth of the price (1,067 points, 955 comments)](https://news.ycombinator.com/item?id=49896586)

---

Canonical: https://www.thedeepfeed.ai/posts/2026-10-09-gpt-6-1-sol-prices-speed-above-astra/
Site: https://www.thedeepfeed.ai
Full corpus: https://www.thedeepfeed.ai/llms-full.txt