GPT-6.1 Sol Costs One-Fifth of Astra at $2 and $10 per Million Tokens, on OpenAI's Own Tests
A.I. / news
GPT-6.1 Sol Costs One-Fifth of Astra at $2 and $10 per Million Tokens, on OpenAI's Own Tests
OpenAI released the model at DevDay on September 29. Every benchmark behind the price claim was run by OpenAI, and the faster Sol tier has not shipped.
OpenAI released GPT-6.1 Sol on September 29 at DevDay 2026, pricing it at $2 per million input tokens and $10 per million output tokens, one-fifth of GPT-6 Astra's $10 and $50. The company says it matches Astra on a software-engineering benchmark at roughly one-fifth of the cost per task.
OpenAI's own announcement page returned a bot-challenge screen when fetched, so the figures below come from VentureBeat, Vellum and MacRumors, plus the title in OpenAI's news feed. It is sold through the API and inside ChatGPT products.
Pricing against Astra
VentureBeat puts cached input for Sol at $0.10 per million tokens, half the prior generation's cached rate. Astra's cached input is $1. Standard uncached input and output are exactly one-fifth of Astra's, the outlet reports.
- GPT-6.1 Sol10 USD
- GPT-6 Astra50 USD
- GPT-6 Astra Ultrafast300 USD
Source: VentureBeat, accessed 2026-09-30
The Ultrafast tier carries a 6x multiplier. Astra Ultrafast is $60 input and $300 output, and VentureBeat derives $12 and $60 for a Sol Ultrafast tier that OpenAI says is coming in the coming days. It was not available on September 30.
The benchmark claims are OpenAI's
VentureBeat notes the figures are OpenAI-run evaluations rather than independent production tests. OpenAI itself acknowledges its evaluations can differ somewhat from production ChatGPT.
Vellum reports 75.2% on DeepSWE v1.1, with task costs about 80% lower than Astra, and 71.4% on OSWorld 2.0, 2.1 percentage points behind Astra at seventh the cost per task. It lists Claude Sonnet 5.5 at 71.0% on DeepSWE v1.1 at the same $2 and $10 pricing.
| Benchmark | GPT-6.1 Sol | Claim against rival |
|---|---|---|
| DeepSWE v1.1 | 75.2% | Matches Astra; Sonnet 5.5 at 71.0% |
| OSWorld 2.0 | 71.4% | 2.1 points behind Astra |
| AutomationBench | 36.0% | 2.2 points above Opus 5.5 |
All three rows are vendor-supplied. No named third party is reported to have reproduced them.
Where it runs and what is missing
Sol is available in the API and in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users. VentureBeat reports it is not yet in regular Chat. The same outlet puts Ultrafast speed at up to 300 tokens per second through the API, against 201 for Google's Gemini 3.5 Flash in Artificial Analysis testing.
OpenAI did not publish parameter counts or training details in the coverage reviewed. VentureBeat's report gives no context window.
The next checkpoint is the Sol Ultrafast release, which OpenAI said would follow within days. Independent runs of DeepSWE and OSWorld would settle whether the one-fifth cost claim holds outside OpenAI's harness. For earlier OpenAI financing, see the $30 billion round at a $1.4 trillion pre-money valuation. For a recent open-weight comparison point, see the Zdtaichu 5.0 9B licence and benchmark report.
Sources
More in A.I.
- 01Qwen3.8-27B Ships Under Apache 2.0 and Fits in 17GB, but Spends 160 Million Tokens Where the Median Spends 43 MillionAlibaba's open-weight model scores 52 on Artificial Analysis's Intelligence Index. Its own benchmark figures are vendor-supplied, and users report slow runs.
- 02OpenAI Ties Moonshot AI to a July Campaign That Replayed Encrypted Reasoning, Offers No Evidence PubliclyOpenAI says 16,000 requests from more than 4,000 accounts tried to recover hidden model reasoning. Its attribution to Moonshot rests on its own assertion.
- 03Gemini 4 Argon Leads 13 of 18 Benchmarks Google Chose, but Only Cyber Defenders Can Use ItGoogle priced the model at $2 and $10 per million tokens and gave access first to its Fairwind Program, with a guardrail-free version for trusted defenders.
- 04Amazon Releases Strands Decider 2B, an Apache 2.0 Decision Model Built on Qwen3.5-2BAWS's Strands Labs scores 72.3 percent on JevBench at a 106 ms median on an RTX 3090. TypeSafe's CEO calls the current crop of rivals less serious than his own team.