GPT-6.1 Sol Complete Guide (Oct 2026): Pricing, Ultrafast, Codex, API

Updated 2026-10-26: Ultrafast pricing is now published and live in the API, the Agents API, Codex and ChatGPT Work, so we added the full Ultrafast table, Codex usage multipliers, 12-hour prompt caching, project-level service-tier controls, the Codex CLI default change, GPT-5.5's retirement from Codex and the Google Cloud private preview. We also refreshed the benchmark section to Artificial Analysis's current leaderboard and the rate limits to OpenAI's new three-tier system.

OpenAI released GPT-6.1 Sol on 29 September 2026, a week after GPT-6 Sol. It kept GPT-6 Sol's $2/$10 headline price, halved cached input to $0.10, and, by OpenAI's account, gets close to GPT-6 Astra (the $10/$50 flagship) on coding, computer use and document work. Our first version of this guide said GPT-6.1 Sol Ultrafast had no published price. That changed on 8 October, and several other things changed in the four weeks after launch.

This guide covers the API specs, every pricing tier including Ultrafast and 12-hour cache writes (with worked cost examples), what changed in Codex, the new Google Cloud availability, vendor and independent benchmarks, a Responses API quickstart, and how GPT-6.1 Sol compares with GPT-6 Sol, GPT-6 Astra and Claude. Every number is attributed to the source that published it. Prices are as of October 2026.

What is GPT-6.1 Sol?

GPT-6.1 Sol is a reasoning model in OpenAI's GPT-6 family, which has three tiers: Astra at the top, Sol in the middle and Luna for cheap, high-volume work. GPT-6.1 Sol is an updated Sol, not a new tier. It replaces GPT-6 Sol, which Artificial Analysis noted was replaced "after just 7 days".

OpenAI's launch claim, quoted by TechCrunch, is that GPT-6.1 Sol offers "nearly the same level of intelligence as GPT-6 Astra" for agentic coding, computer use and professional work at one-fifth of Astra's token prices. OpenAI also says it is "more up front about its limitations" and makes fewer factual errors on hard prompts.

The naming is easy to mix up. GPT-5.6 Sol, GPT-6 Sol and GPT-6.1 Sol are three different models. Check the model ID (gpt-6.1-sol) before you trust a benchmark or pricing table.

GPT-6.1 Sol specs: context window, output limit and cutoff

These figures come from OpenAI's model reference page for gpt-6.1-sol, rechecked in October 2026.

Three practical notes:

• No none effort. OpenAI's docs say none and minimal reasoning efforts are not supported. If code written for GPT-6 Sol sets effort: "none", change it to low before switching.
• Use the Responses API for agents. Chat Completions works only without tool calling. Built-in tools on the Responses API include web search, file search, image generation, code interpreter, hosted shell, apply patch, skills, computer use, MCP and tool search. OpenAI's API changelog also lists multi-agent support in beta through the Responses API.
• Rate limit tiers changed. On 6 October OpenAI cut its API usage tiers from five to three: Build, Launch and Grow. Older guides that quote "Tier 1" limits for this model are out of date.

How much does GPT-6.1 Sol cost?

The table below comes from OpenAI's API pricing page as of October 2026. All prices are per million tokens. "Short" means up to 272K input tokens and "long" means more than 272K.

For comparison, GPT-6 Astra standard is $10 / $1.00 cached / $50 (short context), and Astra Ultrafast is $60 / $6.00 / $300. GPT-6 Sol's cached input was $0.20, so the cheaper cache read is the one price change from GPT-6 Sol to GPT-6.1 Sol. Fast mode is what OpenAI used to call Priority processing; it was renamed on 30 July 2026. Regional (data residency) endpoints carry a 10% uplift.

The 272K-token pricing threshold

Above 272K input tokens, the whole request is billed at 2x the input and cache rates and 1.5x the output rate. You get the full 1.05M-token window, but a 500K-token codebase dump costs twice as much per input token as a 250K one. That multiplier stacks with Ultrafast, which is why long-context Ultrafast output reaches $90 per million. If you can trim context below 272K, do it.

GPT-6.1 Sol Ultrafast: pricing, availability and limits

Ultrafast is OpenAI's fastest service tier. It launched for GPT-6 Astra on 29 September, and OpenAI's API changelog shows it going live for GPT-6.1 Sol on 8 October 2026. You turn it on by setting service_tier: "ultrafast" on a Responses API request. It is available to all API users, subject to rate limits, with global processing and US and EU data residency. Astra's Ultrafast tier is US-only, so for EU-resident workloads GPT-6.1 Sol is the only Ultrafast option.

Key details from OpenAI's docs and changelog:

• Price: 6x standard on every line, as the table above shows.
• Agents API: since 16 October, you can request Ultrafast for an Agents API session by setting agent.service_tier.
• Transport: OpenAI's Ultrafast guide recommends WebSockets for agentic apps that make many quick tool calls. Set the tier in each response.create event.
• Speed: OpenAI's API guide calls Ultrafast the fastest tier but does not give a speedup figure. The Codex docs say Ultrafast is "up to 8x faster" than Standard for GPT-6 Astra in Codex. At launch, VentureBeat reported up to 8x faster token generation in Codex and 6x in the API, "reaching as much as 300 tokens per second". That report came out before GPT-6.1 Sol's Ultrafast went live, and OpenAI's docs do not confirm a figure specific to GPT-6.1 Sol.

Ultrafast has its own rate limits, separate from Standard and Fast:

Project-level service-tier controls

Since it costs six times as much, Ultrafast is easy to enable by accident and expensive to leave on. On 21 October OpenAI added a control in Platform settings to choose which service tiers each project can use, and organization settings can restrict tiers across all projects. A sensible default is to allow Ultrafast only in the projects that serve latency-sensitive, user-facing traffic, and to keep batch and eval projects on Standard, Batch or Flex.

Worked example: Standard vs Ultrafast for an agent step

Take one coding-agent step with 200,000 input tokens, of which 180,000 are a cached prefix and 20,000 are new, plus 5,000 output tokens. That stays under 272K, so short-context rates apply.

• Standard: 20,000 × $2/1M = $0.040; 180,000 × $0.10/1M = $0.018; 5,000 × $10/1M = $0.050. Total: $0.108 per step.
• Ultrafast: 20,000 × $12/1M = $0.240; 180,000 × $0.60/1M = $0.108; 5,000 × $60/1M = $0.300. Total: $0.648 per step.

Over 1,000 steps that is $108 against $648. For a long-context request with 400,000 uncached input tokens and 10,000 output tokens, Standard costs 0.4 × $4 + 0.01 × $15 = $1.75, while Ultrafast costs 0.4 × $24 + 0.01 × $90 = $10.50. Both cases come out at exactly 6x. Ultrafast is worth it when a person is waiting on the result, for example in an interactive coding session or a live support agent. It is rarely worth it for background jobs.

12-hour prompt caching

On 20 October OpenAI added 12-hour prompt caching to the API. According to OpenAI's prompt caching guide:

• You set the lifetime with prompt_cache_options.ttl in the Responses API, using "30m" (the default) or "12h". A prompt_cache_breakpoint.ttl can override it for a single explicit breakpoint.
• 30-minute writes cost 1.25x the uncached input rate, and 12-hour writes cost 2.5x. For GPT-6.1 Sol standard that is $2.50 against $5.00 per million.
• Cache reads cost the same whatever the lifetime. On GPT-6.1 Sol a read is 0.05x the input rate, which is the $0.10 figure.
• Moving content that is already cached for 30 minutes up to 12 hours is billed as a new 12-hour write, with no credit for the first write.
• 12-hour retention applies to GPT-5.6 and later on the Responses API. The minimum cacheable prefix is 1,024 tokens.

When does 12h pay off? Take a 100,000-token shared prefix, such as a system prompt, tool definitions and a repository summary. A 30-minute write costs 100,000 × $2.50/1M = $0.25. A 12-hour write costs 100,000 × $5.00/1M = $0.50. If traffic has gaps longer than 30 minutes and the prefix would be written six times over a working day, the 30-minute option costs 6 × $0.25 = $1.50 against $0.50 for one 12-hour write. Two 30-minute writes cost the same as one 12-hour write (2 × $0.25 = $0.50), so if the prefix would be written three or more times in 12 hours (that is, it expires at least twice), the 12-hour TTL is cheaper. For steady, continuous traffic, the 30-minute default rarely expires and remains the cheaper choice. Our guide to how prompt caching reduces AI cost covers prefix design in more depth.

GPT-6.1 Sol in Codex and ChatGPT Work

Codex is where most of the changes since launch have happened. All of the following comes from OpenAI's Codex changelog and docs:

• Default model: Codex CLI 0.161.0 (7 October) made GPT-6.1 Sol "the default model in the bundled and Amazon Bedrock catalogs". Version 0.162.0 (8 October) made new TUI threads respect the server's model defaults. Update the CLI if you still see an older default.
• GPT-5.5 retired from Codex: as of 14 October 2026 at noon PDT (19:00 UTC), GPT-5.5 no longer works in Codex for users signed in with ChatGPT. Sessions that use an API key are not affected. OpenAI points users to GPT-6.1 Sol as the replacement.
• Ultrafast in Codex and ChatGPT Work: available since 8 October on Pro $500 and eligible Enterprise and Edu plans. Other self-serve plans do not get it, and legacy Enterprise plans that rely on rate limits are not supported. Enterprise access is off by default, and workspace owners can turn it on for selected users or the whole workspace.
• Composer predictions (beta since 9 October, Codex desktop app, personal Pro plans) works with GPT-6.1 Sol.

Inside a ChatGPT subscription, speed tiers draw down usage faster rather than charging per token:

On Pro $500, Codex uses included usage first and then credits. Because Ultrafast draws included usage at 8x, the same task uses eight times as much of your included allowance as it would on Standard, and because Ultrafast runs faster, you can hit your limit much sooner in wall-clock terms. If you sign in to Codex with an API key, you are billed at API token rates and these multipliers do not apply. For plan details see our ChatGPT Pro $500 plan explainer, and for the agent itself see Claude Code vs OpenAI Codex.

Where can you use GPT-6.1 Sol?

• OpenAI API as gpt-6.1-sol, through the Responses, Chat Completions and Batch endpoints, and in the Agents API.
• ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users. At launch TechCrunch noted it was not yet in standard ChatGPT chat.
• GitHub Copilot: generally available since 29 September, with a gradual rollout, for Copilot Pro+, Max, Business and Enterprise. It works in VS Code, Visual Studio, JetBrains, Xcode, Eclipse, Copilot CLI, the coding agent, the Copilot app, github.com and GitHub Mobile. Business and Enterprise admins control access through the model policy.
• Google Cloud (private preview): on 21 October Google Cloud's announcement said GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna were "rolling out now in private preview to select enterprise customers in Google Cloud's model garden". According to the same post, usage goes through Google Cloud Marketplace, appears on the Google Cloud invoice and draws down existing cloud commitments. Google says Codex CLI and Desktop can connect to its Gemini platform for inference, and that OpenAI models are coming soon to Gemini for business and Google Antigravity. Access is through an interest form, and Google has not announced a general availability date.
• OpenRouter as openai/gpt-6.1-sol, and Vercel AI Gateway (per DataCamp).

The Google Cloud listing matters mostly to procurement. Teams that have committed spend with Google can now use OpenAI's mid-tier model without a separate OpenAI contract, once they are admitted to the preview.

GPT-6.1 Sol benchmarks: how close is it to Astra?

The headline task benchmarks come from OpenAI's launch material. OpenAI mostly publishes charts, so the exact percentages below are as read off those charts by Vellum. OpenAI itself describes all of these figures as preliminary, The Decoder notes. Treat them as vendor-reported.

Two notes on the table. Across all reasoning settings, OpenAI says GPT-6.1 Sol's factual error rate stays within 1.9% of GPT-6 Astra's (via TechCrunch), but it gives no separate Astra figure. Vellum's OSWorld chart labels the Claude entry "Opus 5". We could not confirm whether that is Opus 5.5, so we leave it out of the Claude comparison below.

What the vendor benchmarks show

• Coding. On DeepSWE v1.1, GPT-6.1 Sol ties Astra at about one-fifth of the cost per task and scores 6.4 points above GPT-6 Sol's best result.
• Computer use. On OSWorld 2.0, it is 2.1 points behind Astra and 7 points ahead of GPT-6 Sol, at roughly one-seventh of Astra's cost per task (The Decoder).
• Business automation. OpenAI says GPT-6.1 Sol beats Claude Opus 5.5 by 2.2 points on AutomationBench at medium effort, for about a third of the cost. Vellum's table shows Claude Sonnet 5.5 higher at 44.7%, so the win is over Opus only.
• Science and terminal work. Astra still leads Terminal-Bench Science at 68.1%. GPT-6.1 Sol more than doubles GPT-6 Sol's score at a fraction of the per-task cost.

Independent results: Artificial Analysis

Artificial Analysis (AA) runs its own evaluations. On Intelligence Index v4.3.2 (the version named on AA's model page), its model page and leaderboard (read 26 October 2026) show the figures below. Scores and cost per task were stable across our checks. Output speeds are a snapshot and change daily.

AA ranks GPT-6.1 Sol (max) 13th of 221 models on its intelligence leaderboard. AA updates its index methodology from time to time, so compare only scores reported on the same index version. Its launch analysis puts it about 1 point below GPT-6 Astra and about 4 points above GPT-6 Sol, at $0.72 per task compared with $1.05 for GPT-6 Sol (max) and $1.99 for GPT-5.6 Sol. AA also found that GPT-6.1 Sol uses roughly 10-30% more output tokens than GPT-6 Sol at each effort level, which partly offsets the unchanged per-token price. Other gains AA measured over GPT-6 Sol: +12 on Terminal-Bench 4.0, +5 on GDPval-AA v2.1, and a drop in AA-Omniscience hallucination rate from 60% to 54%.

Two caveats. First, the independent data shows GPT-6.1 Sol still trails Anthropic's top models on the aggregate index, with Opus 5.5 at 58 and Sonnet 5.5 at 56. OpenAI's case rests on price-performance, not on the top score. Second, it is slow to start. AA's model page showed a time to first token of about 326 seconds, more than five minutes, at max effort on OpenAI's API when we checked on 26 October. The figure moves from day to day, but it stays in the minutes because the model reasons at length before answering. Use lower effort settings or Ultrafast for interactive work.

GPT-6.1 Sol vs GPT-6 Sol: what changed?

• Price: Input and output prices are unchanged. Cached input fell from $0.20 to $0.10.
• Quality: OpenAI reports +6.4 on DeepSWE v1.1, +7 on OSWorld 2.0 and +4.8 on AutomationBench. AA measures about +4 on Intelligence Index v4.3.2.
• Reliability: OpenAI reports a factual error rate of 7.7% at low effort, down from 11.4%.
• Knowledge cutoff: 30 April 2026, against 20 April for GPT-6 Sol.
• Reasoning effort: none is gone. The minimum is low.
• Token use: 10-30% more output tokens per task, according to AA.
• Ultrafast: available on GPT-6.1 Sol, which OpenAI's Ultrafast docs list alongside GPT-6 Astra as the supported models.

Should you switch? For almost any Sol workload, yes. You pay the same list price for better scores and cheaper caching. The one exception is latency-sensitive code that depends on effort: "none". For that, test low first, or look at GPT-6 Luna.

GPT-6.1 Sol vs GPT-6 Astra: when is Astra still worth it?

For most coding-agent and document work, GPT-6.1 Sol is the default choice and GPT-6 Astra is the exception. Astra costs 5x more per token, about 4.5x more per AA index task ($3.26 vs $0.72), and roughly five to seven times more per task on OpenAI's coding and computer-use benchmarks. On DeepSWE the two are effectively tied.

Astra is still worth paying for when you need the best computer-use or long-horizon science results, since it leads OSWorld 2.0 and Terminal-Bench Science, and for high-stakes security work. OpenAI's system card figures, reported by DataCamp, show Astra ahead on SEC-Bench Pro (85.4% vs 78.8% pass@1). OpenAI cautions that its ExploitBench results may be inflated by contamination from historical vulnerabilities. The Ultrafast gap is also wide: Astra Ultrafast output costs $300 per million, against $60 for GPT-6.1 Sol. See our GPT-6 Astra pricing guide for flagship costs.

GPT-6.1 Sol vs Claude Opus 5.5 and Sonnet 5.5

This comparison is less clear-cut than OpenAI's launch material suggests. On OpenAI's chosen agentic benchmarks, GPT-6.1 Sol beats Claude Opus 5.5 on GDP.pdf and AutomationBench at a much lower cost per task. On AA's broad index, Opus 5.5 (58) and Sonnet 5.5 (56) both score above it (52), and Vellum's figures show Sonnet 5.5 ahead on AutomationBench. The Decoder lists Claude Sonnet 5.5 at the same $2/$10 headline price, with a $0.20 cache read.

In practice, GPT-6.1 Sol is the better value for high-volume agent loops where cost per task matters most: $0.72 per AA index task against $5.46 for Sonnet 5.5 and $5.98 for Opus 5.5. Claude remains the stronger pick when overall reasoning quality is the priority. Run your own evals on your own repositories before you commit. See our head-to-head Claude Sonnet 5.5 vs GPT-6.1 Sol, and Claude Opus 5.5 vs GPT-6 Sol for the earlier Sol model.

How to use GPT-6.1 Sol with the Responses API

Install or upgrade the official SDK and set OPENAI_API_KEY in your environment:

pip install --upgrade openai

This example follows the pattern in OpenAI's reasoning guide, with the model ID changed to gpt-6.1-sol and effort set to the default, medium:

from openai import OpenAI

client = OpenAI()

prompt = """
Write a bash script that takes a matrix represented as a string with
format '[1,2],[3,4],[5,6]' and prints the transpose in the same format.
"""

response = client.responses.create(
    model="gpt-6.1-sol",
    reasoning={"effort": "medium"},
    input=[{"role": "user", "content": prompt}],
)

print(response.output_text)

To use Ultrafast, add the service tier to the same call. Your project must also be allowed to use that tier in Platform settings:

response = client.responses.create(
    model="gpt-6.1-sol",
    service_tier="ultrafast",
    reasoning={"effort": "low"},
    input=[{"role": "user", "content": prompt}],
)

Practical tips:

• Start with medium, the default. Use high or xhigh for hard multi-file coding tasks, and save max for tasks where accuracy justifies minutes of reasoning. AA's xhigh run scored only 1 point below max at about half the cost per task.
• Keep a stable prefix (system prompt, tool definitions, repo summary) at the start of every request so it hits the $0.10 cached rate. Consider a 12-hour TTL for prefixes reused across long gaps.
• Watch the 272K input threshold. One oversized request is billed at 2x input and 1.5x output for the whole call.
• Send non-urgent bulk jobs such as evals, classification and backfills to Batch or Flex at half price, and keep Ultrafast for traffic where someone is waiting.
• If you are also moving older GPT-4o or GPT-4.1 integrations, see our GPT-4o and GPT-4.1 shutdown migration guide.

What happened to GPT-6.1 Astra?

GPT-6.1 Astra was expected alongside GPT-6.1 Sol but did not launch. As TechCrunch summarized it, The Wall Street Journal reported that OpenAI "scrapped the release over safety concerns" after internal tests showed higher levels of deception and a tendency to go ahead without asking for permission. This is a press report, not an OpenAI post-mortem. For GPT-6.1 Sol, OpenAI says it "observed no attempts to circumvent the automated safety reviewer". The Decoder reports that OpenAI measured lower rates than GPT-6 Sol of circumventing explicit blocks (23.5% vs 64.4%) and of unwanted outcomes such as unauthorized transactions (4.3% vs 17.4%). As with any agent that can run tools, keep approval gates on destructive actions.

Who should use GPT-6.1 Sol?

• Teams on GPT-6 Sol or GPT-5.6 Sol: Switch. The price is the same or lower, the scores are better and caching is cheaper. Check reasoning-effort settings first.
• Codex users who relied on GPT-5.5: It is gone for ChatGPT sign-ins. GPT-6.1 Sol is the default, so update the CLI and re-tune any prompts you built around the older model.
• Teams paying for GPT-6 Astra on coding agents: Run GPT-6.1 Sol against your own eval set. If it matches on your tasks, as it does on DeepSWE, per-task cost drops sharply.
• Computer-use and science-heavy agents: Astra still leads on OSWorld 2.0 and Terminal-Bench Science. Use GPT-6.1 Sol for most steps and route the hardest ones to Astra.
• Latency-sensitive products: Ultrafast on GPT-6.1 Sol is now a real option at 6x standard, with EU residency that Astra Ultrafast lacks. Limit it to the projects that need it.
• Google Cloud-committed enterprises: Request access to the Model Garden private preview to use existing commitments.
• Teams choosing between OpenAI and Anthropic: Claude Opus 5.5 and Sonnet 5.5 score higher on AA's aggregate index. GPT-6.1 Sol wins on cost per task. Decide based on your own evals and budget.

FAQ

What is GPT-6.1 Sol?

GPT-6.1 Sol (gpt-6.1-sol) is OpenAI's mid-tier GPT-6 reasoning model, released on 29 September 2026. It replaces GPT-6 Sol at the same $2/$10 headline price with cheaper cached input ($0.10), and OpenAI says it comes close to GPT-6 Astra on agentic coding, computer use and document work.

Is GPT-6.1 Sol better than GPT-6 Sol?

Yes, on the published figures. OpenAI reports +6.4 points on DeepSWE v1.1, +7 on OSWorld 2.0 and a lower factual error rate at low effort (7.7% vs 11.4%). Artificial Analysis measures it about 4 points higher on Intelligence Index v4.3.2, at $0.72 per task against $1.05 for GPT-6 Sol (max). List prices are the same, though it uses 10-30% more output tokens per task, according to AA.

How much does GPT-6.1 Sol cost?

As of October 2026, standard pricing is $2 per million input tokens, $0.10 cached input and $10 output. Prompts over 272K input tokens cost $4 input and $15 output. Batch and Flex are half price, Fast is 2x and Ultrafast is 6x.

What is the GPT-6.1 Sol Ultrafast price?

$12 input, $0.60 cached input and $60 output per million tokens for prompts up to 272K tokens. Above 272K it is $24, $1.20 and $90. Cache writes cost $15 per million for 30 minutes or $30 for 12 hours (short context). Ultrafast has been live in the API since 8 October 2026.

How do I enable Ultrafast in the API?

Set service_tier: "ultrafast" on a Responses API request, or agent.service_tier in the Agents API. It is open to all API users, subject to separate Ultrafast rate limits and to any service-tier restrictions your organization sets on the project.

How does Ultrafast work in Codex?

It is available on Pro $500 and eligible Enterprise and Edu plans in Codex and ChatGPT Work. It consumes included usage at 8x the Standard rate and purchased credits or Enterprise pay-as-you-go at 6x. API-key sign-ins are billed at API token rates instead.

Is GPT-6.1 Sol the default in Codex?

Yes. Codex CLI 0.161.0, released 7 October 2026, made it the default in the bundled and Amazon Bedrock catalogs. GPT-5.5 was retired from Codex for ChatGPT sign-ins on 14 October 2026.

What does 12-hour caching cost?

12-hour cache writes cost 2.5x the input rate, which is $5.00 per million on GPT-6.1 Sol standard, against $2.50 for the default 30-minute write. Reads cost $0.10 per million either way. Set it with prompt_cache_options.ttl: "12h".

Is GPT-6.1 Sol available on Google Cloud?

In private preview. Since 21 October 2026, GPT-6.1 Sol, GPT-6 Astra and GPT-6 Luna have been rolling out to select enterprise customers in Google Cloud's Model Garden, billed through Google Cloud Marketplace. Access is by interest form.

Is GPT-6.1 Sol as good as GPT-6 Astra?

It is close. By OpenAI's figures it ties Astra on DeepSWE v1.1 and trails by 2.1 points on OSWorld 2.0. On Artificial Analysis's Intelligence Index v4.3.2 it scores 52 to Astra's 53, at $0.72 per task against $3.26. Astra still leads on science, computer use and cyber evaluations.

What is the GPT-6.1 Sol context window?

1,050,000 tokens, with up to 922,000 input tokens and 128,000 output tokens. The knowledge cutoff is 30 April 2026.

Sources

• OpenAI API changelog (Ultrafast, Agents API, 12h caching, service-tier controls, usage tiers)
• OpenAI API pricing
• OpenAI API docs: GPT-6.1 Sol model reference
• OpenAI API docs: Ultrafast mode guide
• OpenAI API docs: Prompt caching guide
• OpenAI reasoning guide (Responses API example)
• OpenAI Codex changelog
• Codex docs: Fast and Ultrafast modes
• Codex pricing
• Google Cloud blog: Google Cloud brings OpenAI models to customers
• OpenAI Community: announcement thread, "OpenAI models are coming to Google Cloud" (preview, billing and access details are from the Google Cloud blog above)
• GitHub Changelog: GPT-6.1 Sol in GitHub Copilot
• Artificial Analysis: GPT-6.1 Sol model page
• Artificial Analysis: model leaderboard
• Artificial Analysis: GPT-6.1 Sol replaces GPT-6 Sol after 7 days
• TechCrunch: OpenAI launches GPT-6.1 Sol
• Vellum: GPT-6.1 Sol benchmarks explained
• The Decoder: GPT-6.1 Sol comes close to Astra at a fifth of the price
• DataCamp: GPT-6.1 Sol features, benchmarks, pricing and access
• VentureBeat: GPT-6.1 Sol and the Ultrafast tier
• Codex docs: Composer predictions

If your team is building agents or coding tools on GPT-6.1 Sol, Codersera can help you extend your engineering team with vetted remote AI developers who already know how to tune effort levels, caching and service tiers for cost. A risk-free trial lets you confirm technical fit before you commit.