GPT-6.1 Sol: API Pricing, Specifications and How to Use It

What is GPT-6.1 Sol?

GPT-6.1 Sol is the updated Sol-tier model in OpenAI's GPT-6 family. Its positioning is stronger agentic coding, computer use and professional work, with capabilities approaching Astra at a lower price. That is a reason to evaluate it, not proof that every task produces Astra-level results.

For a new project, test GPT-6.1 Sol on representative repository tasks, browser workflows and professional documents. For an existing Sol deployment, compare the models on identical prompts, tools and acceptance tests before changing the default. An old gpt-6-sol alias does not automatically mean gpt-6.1-sol.

GPT-6.1 Sol specifications

Specification	Published value or qualification
API model ID	gpt-6.1-sol
Context window	1,050,000 tokens
Maximum output	128,000 tokens
Approximate input budget with the full output allowance reserved	922,000 tokens; arithmetic budget, not a separately guaranteed input limit
Knowledge cutoff	April 30, 2026
Input modalities	Text and images
Output modality	Text
Reasoning efforts	low, medium (default), high, xhigh, max
Unsupported reasoning settings	none, minimal
Parameter count	Not publicly disclosed; closed weights

The context window is a shared budget, not 1,050,000 input tokens plus 128,000 output tokens. Instructions, tools, conversation history and output all consume that budget; a provider may impose a smaller limit. Reasoning tokens also consume output budget. Set an explicit effort such as medium for a first comparison rather than relying on different clients' defaults.

GPT-6.1 Sol API pricing and cache costs

Standard OpenAI rates, in US dollars per million tokens:

Billing category	GPT-6.1 Sol	GPT-6 Astra
Uncached input	$2	$10
Cached input	$0.10	$1
Output	$10	$50
Cache write	$2.50	$12.50

At $0.10, a cache hit costs 95% less than Sol 6.1's standard input. Caching is therefore relevant to agents that repeatedly send stable instructions and repository context. A cache write is not a cache hit; do not assume every input token gets the cached rate or that third-party routes copy OpenAI's policy.

Long-context pricing matters: OpenAI documents higher rates for Sol 6.1 and Astra when input exceeds 272K tokens: 2× input, cache-hit and cache-write rates, and 1.5× output pricing, applied to the full request. For Sol 6.1, that means $4 input / $0.20 cached input / $5 cache writes / $15 output per million tokens. Check OpenAI pricing and your route's billing conditions.

For illustration, 100,000 uncached input tokens and 20,000 output tokens cost $0.40 on Sol 6.1 versus $2.00 on Astra at standard rates, excluding tool fees. This assumes identical billable token counts. Different reasoning lengths, retries and tool calls mean a one-fifth token price does not guarantee a one-fifth task cost.

How to use GPT-6.1 Sol in ChatGPT Work and Codex

OpenAI's model availability page lists ChatGPT Work and Codex access on Plus, Pro, Business, Enterprise and Edu. Enterprise and Edu keep Sol 6.1 off by default until an administrator enables it; Free and Go are not included at launch. Check the picker in Work or Codex, or select it in the CLI:

codex -m gpt-6.1-sol

If it is missing, check the client version and workspace policy. Standard and Fast modes are available at launch; Ultrafast is a later rollout.

Ordinary ChatGPT Chat is a separate surface. Work and Codex availability does not establish ordinary Chat availability. A ChatGPT subscription also does not automatically include pay-as-you-go API credit.

OpenAI's September 29 Codex changelog also records adding Sol 6.1 to the Amazon Bedrock Mantle and Runtime catalogs. For GitHub Copilot, OpenRouter or Devin, consult that product's own catalog. Subscription access, regions, aliases and API identifiers can differ; verify the integration rather than assuming that an OpenAI identifier works everywhere.

GPT-6.1 Sol API example

For the official API, use an OpenAI API key and the official base URL. For APIMaster, register, create an API key, and check the GPT-6.1 Sol route card. This request template uses https://apimaster.ai/v1 and gpt-6.1-sol:

export APIMASTER_API_KEY="your-api-key"

curl "https://apimaster.ai/v1/responses" \
  -H "Authorization: Bearer $APIMASTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-6.1-sol",
    "input": "Review this Python function for edge cases and propose tests: def mean(values): return sum(values) / len(values)",
    "reasoning": {"effort": "medium"},
    "max_output_tokens": 2048
  }'

This is a template, not a successful-call benchmark. Confirm that the selected route supports Responses API and the reasoning setting. OpenAI documents tool calling through Responses; GPT-6.1 Sol supports Chat Completions without tool calling. On an access error, check the key and permissions rather than silently falling back to another model.

GPT-6.1 Sol vs GPT-6 Sol vs GPT-6 Astra

Decision	GPT-6.1 Sol	GPT-6 Sol	GPT-6 Astra
Standard input / output price	$2 / $10	$2 / $10	$10 / $50
Cached input (per 1M)	$0.10	$0.20	$1
Evaluation role	Upgraded Sol candidate	Existing Sol baseline	Higher-price comparison
Reason to choose it	Agent workflows balancing cost and quality	Already validated workloads	Tasks where tests justify the extra cost
What not to assume	Equal performance to Astra on every task	An old alias automatically upgrades	Higher price guarantees better results

Measure completed tasks, failing tests, factual mistakes, latency and total billed cost, with identical tool permissions and stopping rules. Inspect code changes and run tests; a persuasive explanation is not a correct patch.

Early impressions that Sol feels close to Astra are not an independently reproduced ranking. Unsourced Intelligence Index scores, DeepSWE results and OSWorld gaps are not treated as verified numbers here. Check the dated original report, reasoning effort and test setup before using a benchmark to choose a provider.

What is GPT-6.1 Sol Ultrafast?

GPT-6.1 Sol Ultrafast is coming later, according to OpenAI's model availability page. Standard and Fast are the launch options. There is no confirmed Sol 6.1 Ultrafast speed guarantee in the documentation checked for this article.

The published up to 8× token-generation speed comparison currently refers to GPT-6 Astra Ultrafast versus Astra Standard in Codex, not Sol 6.1. See OpenAI's speed documentation. Do not transfer Astra's speed claim or plan eligibility to a future Sol mode.

Generation speed is not end-to-end task speed: tools, tests, network latency and retries still take time. Do not invent an Ultrafast API model ID or assume that the standard model's API price applies to the new mode.

FAQ

Is GPT-6.1 Sol released?

Yes. OpenAI's release notes date the launch to September 29, 2026. This article was checked on September 30, 2026; it is a launch guide, not a rumor tracker.

What is the GPT-6.1 Sol API model name?

Use gpt-6.1-sol on an endpoint that supports it. gpt-6-sol and gpt-6-astra are different IDs. A returned model label alone is not independent proof of model identity.

How much does GPT-6.1 Sol cost?

Standard OpenAI rates are $2 input, $0.10 cached input and $10 output per million tokens. Long-context and service-tier conditions can change the bill. On September 30, APIMaster had 3 active routes from $0.10707 / $0.53535 input / output; the live card is authoritative.

Is GPT-6.1 Sol available in ordinary ChatGPT Chat?

Not at launch. GPT-6.1 Sol is available in ChatGPT Work, Codex and the API, but not ordinary ChatGPT Chat. Check the current model picker and official release notes for subsequent changes.

How many parameters does GPT-6.1 Sol have?

OpenAI has not publicly disclosed the parameter count. Context size and benchmark scores do not reveal an exact number of parameters.

Should I replace Astra with GPT-6.1 Sol?

Test Sol 6.1 on your own coding and professional tasks first. Standard input and output rates are one-fifth of Astra's, but quality, retries and reasoning length determine the actual trade-off.

Sources and related guides

• OpenAI model reference: identifier, specifications and reasoning settings.
• OpenAI API pricing: token rates and long-context billing.
• ChatGPT Work changelog: September 29 release and access.
• OpenAI model availability: selection, plan controls and the later Sol Ultrafast rollout.
• OpenAI speed documentation: the Astra-only 8× comparison.
• APIMaster live route data: active routes checked September 30, 2026.
• GPT-6 Sol API guide, Sol vs Astra, and Codex setup.

Start with a verified route and a small workload. Register with APIMaster, create a key at API Keys, then use https://apimaster.ai/v1 with gpt-6.1-sol. Check the live discount card before buying: September 30 routes offered up to ~94.6% off standard pricing, not a permanent price promise.