GPT-6.1 Sol pricing, model id, and Sol vs Astra in Codex
This research library uses AI-assisted source research and drafting. Linked sources support product claims; analysis and proposed exercises are our interpretation. Unless an article documents a test and its results, do not read it as a hands-on review or an independently verified benchmark.
GPT-6.1 Sol is the upgrade to GPT-6 Sol that OpenAI released at DevDay on 29 September 2026. The API model id is gpt-6.1-sol, and standard pricing is $2 per million input tokens, $0.10 per million cached input tokens, and $10 per million output tokens (OpenAI launch post). OpenAI positions it as near-Astra intelligence at a fifth of GPT-6 Astra's standard input and output prices.
How much does GPT-6.1 Sol cost?
These are the standard short-context API prices per million tokens from the OpenAI pricing page. Short context means up to 272K input tokens.
| Model | Input | Cached input | Cache writes | Output |
|---|---|---|---|---|
| gpt-6.1-sol | $2.00 | $0.10 | $2.50 | $10.00 |
| gpt-6-sol | $2.00 | $0.20 | $2.50 | $10.00 |
| gpt-6-astra | $10.00 | $1.00 | $12.50 | $50.00 |
| gpt-6-luna | $0.10 | $0.01 | $0.125 | $0.50 |
The only price change against GPT-6 Sol is cached input, which drops by half. Long-context requests above 272K input tokens cost $4 input, $0.20 cached, $5 cache writes, and $15 output for GPT-6.1 Sol. Batch and Flex halve the standard rates. Fast mode doubles them.
In Codex with ChatGPT credits, GPT-6.1 Sol costs 50 credits per million input tokens, 2.5 per million cached, and 250 per million output. GPT-6 Astra costs 250, 25, and 1,250.
gpt-6.1-sol API specs
The model page lists:
- Context window of 1,050,000 tokens and 128,000 max output tokens.
- Reasoning effort
low,medium(default),high,xhigh, andmax. - Knowledge cutoff of 30 April 2026.
- Chat Completions, Responses, and Batch endpoints.
- Tools including web search, file search, code interpreter, hosted shell, apply_patch, skills, computer use, MCP, and tool search.
- Tier 1 limits of 500 requests and 500,000 tokens per minute.
GPT-6.1 Sol vs GPT-6 Sol and GPT-6 Astra on benchmarks
All numbers below are OpenAI's own, from the launch post. Treat them as a shortlist of what to test, not as a verdict on your codebase.
| Benchmark | What OpenAI reports for GPT-6.1 Sol |
|---|---|
| DeepSWE v1.1 | Matches GPT-6 Astra at roughly a fifth of the cost; 6.4 points above GPT-6 Sol's best score |
| OSWorld 2.0 offline | 7 points above GPT-6 Sol at max effort; within 2.1 points of Astra at about a seventh of the cost per task |
| AutomationBench 1.0.6 | Up 4.8 points from GPT-6 Sol |
| Terminal-Bench Science 0.1 | More than double GPT-6 Sol's score at max effort; $5.47 per task against $23.80 for Astra |
| GDP.pdf | Approaches Astra at about a fifth of the cost per task |
Two caveats matter. Astra still has the highest Terminal-Bench Science score at 68.1%. And OpenAI reports the share of low-effort answers with a factual error falling from 11.4% to 7.7%, which is better but not zero.
A worked cost example with cached input
Take a Codex-style agent loop through the API. It makes 20 requests. Each request sends a 50K-token prefix that never changes, such as repo instructions and file context, plus 10K new tokens. Each response is 2K tokens.
Assume the first request writes the 50K prefix to the cache and the next 19 read it. Prompt caching is on by default. Cache writes cost 1.25 times the input rate.
| Line item | Tokens | gpt-6.1-sol | gpt-6-sol | gpt-6-astra |
|---|---|---|---|---|
| Cache write | 50K | $0.125 | $0.125 | $0.625 |
| Cached reads | 950K | $0.095 | $0.19 | $0.95 |
| New input | 200K | $0.40 | $0.40 | $2.00 |
| Output | 40K | $0.40 | $0.40 | $2.00 |
| Total | $1.02 | $1.115 | $5.575 |
Without any cache hits, the same GPT-6.1 Sol loop costs $2.80. The cached prefix saves $1.78 over 20 requests. The cheaper cache read barely moves the Sol vs GPT-6 Sol total here. It matters more when the stable prefix is large and the loop is long.
The cached prefix stays eligible for reuse for 30 minutes after its last write or read on GPT-5.6 and later models. Keep the stable part of your prompt at the start and the changing part at the end.
When to pick Sol vs Astra in Codex
OpenAI's Codex models guide recommends GPT-6.1 Sol for repeated, long-running work across code, apps, and documents. It recommends Astra as the most capable model for complex work. In practice that gives a simple split:
- Default to GPT-6.1 Sol for feature work, refactors, test writing, and review passes.
- Switch to Astra when a task needs long, sustained reasoning across many systems, or when Sol has failed on it twice.
- Use GPT-6 Luna for narrow, high-volume jobs such as summaries and extraction.
Usage limits point the same way. On Plus and Standard Business, the Codex pricing page estimates 15 to 160 local messages per five hours on GPT-6.1 Sol, against 5 to 45 on Astra.
To switch, type /model in an interactive Codex CLI session, or launch with the flag:
codex --model gpt-6.1-sol
codex exec -m gpt-6.1-sol "Review the current changes"
To make it the default and send /review to Astra, set both in config.toml:
model = "gpt-6.1-sol"
review_model = "gpt-6-astra"
GPT-6.1 Sol is available today in ChatGPT Work and Codex on Plus, Pro, Business, Enterprise, and Edu, and to all API users. It is not yet available in ChatGPT's chat mode. Codex supports it in Standard and Fast modes; GPT-6.1 Sol Ultrafast is listed as coming soon.
Plan for one more deadline. GPT-5.5 retires from ChatGPT, ChatGPT Work, and Codex on 14 October 2026. Replace it in saved settings, scripts, and scheduled tasks before then. The API is not affected.
Keep your repo rules in AGENTS.md so both models get the same instructions, and apply the same review standard to both. That is the core of our Delegate, Review, Own methodology.
Try this on your next five Codex tasks: run them on gpt-6.1-sol at medium effort and note which ones you had to redo on Astra.