📊 See GPT-6.1 Sol rankings

📱 Get the AI Rank app

GPT-6 is a family of three models rather than one. GPT-6 Astra is OpenAI's top model at $10 per million input tokens and $50 per million output tokens. GPT-6 Sol is the mid tier, now on version 6.1 at $2/$10. GPT-6 Luna is the high-volume tier at $0.10/$0.50. All three read text and images, write text, and share a 1,050,000-token context window with up to 128,000 output tokens.

OpenAI released Astra on September 3, 2026, Sol and Luna on September 22, and GPT-6.1 Sol on September 29. In the API they are gpt-6-astra, gpt-6.1-sol (the older gpt-6-sol is still listed) and gpt-6-luna.

GPT-6 Astra vs Sol vs Luna at a Glance

GPT-6 Astra GPT-6.1 Sol GPT-6 Luna
OpenAI's description Most capable, for the hardest end-to-end work Near-Astra performance at a lower cost Most efficient, for focused high-volume tasks
API model ID gpt-6-astra gpt-6.1-sol gpt-6-luna
Released Sep 3, 2026 Sep 29, 2026 (GPT-6 Sol: Sep 22) Sep 22, 2026
Standard input / output (per 1M) $10 / $50 $2 / $10 $0.10 / $0.50
Reasoning effort low to max (no none) low to max (no none or minimal) none to max, default medium
Context / max output 1,050,000 / 128,000 1,050,000 / 128,000 1,050,000 / 128,000
Knowledge cutoff Apr 30, 2026 Apr 30, 2026 May 18, 2026
Ultrafast mode Yes, US residency only Yes, US and EU residency Not listed

The reasoning settings matter in practice. Only Luna accepts reasoning.effort: "none", so it is the one to pick for quick, non-thinking calls. OpenAI's model page for Luna also notes that Chat Completions supports function calling with Luna only when effort is set to none; for tools at other effort levels, use the Responses API. GPT-6.1 Sol supports tool calling through the Responses API.

GPT-6 Pricing by Model and Processing Tier

All prices are USD per million tokens from OpenAI's API pricing page, for prompts up to 272K input tokens.

Model Tier Input Cached input Output 30-min cache write 12-hour cache write
GPT-6 Astra Standard $10.00 $1.00 $50.00 $12.50 $25.00
GPT-6 Astra Batch $5.00 $0.50 $25.00 $6.25 $12.50
GPT-6 Astra Fast $20.00 $2.00 $100.00 $25.00 $50.00
GPT-6 Astra Ultrafast $60.00 $6.00 $300.00 $75.00 $150.00
GPT-6.1 Sol Standard $2.00 $0.10 $10.00 $2.50 $5.00
GPT-6.1 Sol Batch $1.00 $0.05 $5.00 $1.25 $2.50
GPT-6.1 Sol Fast $4.00 $0.20 $20.00 $5.00 $10.00
GPT-6.1 Sol Ultrafast $12.00 $0.60 $60.00 $15.00 $30.00
GPT-6 Luna Standard $0.10 $0.01 $0.50 $0.125 $0.25
GPT-6 Luna Batch $0.05 $0.005 $0.25 $0.0625 $0.125
GPT-6 Luna Fast $0.20 $0.02 $1.00 $0.25 $0.50

Batch and Flex cost half of Standard, Fast costs double, and Ultrafast costs six times Standard. Fast mode is the old Priority processing tier, renamed on July 30, 2026. Data-residency and FedRAMP endpoints add 10%.

Long prompts change the bill. Once a request has more than 272K input tokens, the whole request is billed at 2x the input and cache rates and 1.5x the output rate. That applies to all three models.

Example request (Standard) GPT-6 Astra GPT-6.1 Sol GPT-6 Luna
100K input + 10K output, no cache $1.50 $0.30 $0.015
400K input + 20K output, no cache (long-prompt rates) $9.50 $1.90 $0.095

Caching is where GPT-6.1 Sol moved most. Its cached input is $0.10, half of GPT-6 Sol's $0.20 and 95% below its own standard input rate. GPT-6 Sol keeps the same $2/$10 base rates with the older $0.20 cache price, so there is little reason to start new work on it.

What Each GPT-6 Model Is Built For

GPT-6 Astra. OpenAI's launch post calls it its best model for computer use, coding and professional work, and reports 57.9% on Terminal-Bench 4.0, 74.1% on DeepSWE v1.1, 72.6% on the OSWorld 2.0 offline set and 96.0% on GPQA Diamond, each as the maximum score at any effort setting. Astra reached the Critical cybersecurity threshold in OpenAI's Preparedness Framework, so it refuses some advanced security work, such as writing proof-of-concept exploits, and API tasks can stop when a safety check triggers. Pro, Business and Enterprise ChatGPT plans also get GPT-6 Astra Pro.

GPT-6.1 Sol. OpenAI says it nearly matches Astra on agentic coding, computer use and professional work at one-fifth of Astra's token prices. On DeepSWE v1.1 OpenAI reports it matching Astra at about one-fifth of the cost per task. On the OSWorld 2.0 offline set it comes within 2.1 points of Astra at maximum effort. On Terminal-Bench Science 0.1 it averaged $5.47 per task against $23.80 for Astra, while Astra kept the top score at 68.1%. It also supports Multi-agent in beta, which lets the model hand work to subagents within one Responses API request.

GPT-6 Luna. OpenAI pitches Luna for fast, efficient everyday work at scale. At max effort it scored 66.6% on DeepSWE v1.1 in OpenAI's Sol and Luna announcement. OpenAI used it to launch the Decisions API beta on October 6, 2026, which turns text and images into typed answers. In ChatGPT, Free and Go users can reach Luna in the desktop app.

All benchmark numbers above are OpenAI's own and were run in its research environment.

GPT-6 Luna and Sol on Arena and OpenRouter

AI Rank's Arena data published October 15, 2026 lists each model at maximum reasoning effort:

Model (max effort) Overall rank Overall score (95% CI) Coding rank Coding score (95% CI)
GPT-6.1 Sol 22 1483.8 (1475.7–1492.0) 10 1539.8 (1522.6–1556.9)
GPT-6 Astra 30 1477.3 (1471.0–1483.6) 9 1543.4 (1531.8–1555.0)
GPT-6 Sol 61 1459.0 (1452.7–1465.4) 39 1518.2 (1506.6–1529.8)
GPT-6 Luna 91 1442.8 (1438.2–1447.4) 65 1506.4 (1499.8–1513.1)

GPT-6.1 Sol and Astra have overlapping intervals on both boards, so Arena does not separate them. Both sit clearly above GPT-6 Sol and Luna. Luna has the most votes of the four (49,617 overall), so its position is the most stable.

→ Track it on AI Rank: GPT-6 Astra (max): Arena rating, rank and 90-day history

Usage tells a different story. On OpenRouter's weekly text ranking for October 10–16, 2026, GPT-6 Luna was rank 4 with 5.63T tokens (down 6% week over week). GPT-6.1 Sol was rank 14 with 1.72T (up 9%). GPT-6 Sol fell to rank 32 with 415B (down 49%) as traffic moved to 6.1. GPT-6 Astra was rank 45 with 319B (down 32%). Usage on one router reflects price and workload mix, not quality.

→ Live usage: See where all three rank this week in the OpenRouter model usage ranking on AI Rank (tokens, Oct 10–16, 2026 window).

GPT-6.1 Sol Against Claude Sonnet 5.5

GPT-6.1 Sol and Claude Sonnet 5.5 have the same $2/$10 standard rates. The difference is past 272K input tokens: Sol's whole request moves to long-prompt rates, while Sonnet 5.5 keeps its standard price across its 1M window. On Arena, GPT-6.1 Sol max leads Sonnet 5.5 xhigh (1474.9 overall, rank 36) but the intervals overlap. The full comparison, including Bedrock and cache details, is in Claude Sonnet 5.5: pricing, benchmarks and GPT-6.1 Sol comparison.

Choosing Between Astra, Sol and Luna

  • Default to GPT-6.1 Sol for coding agents, computer use and document work. It costs a fifth of Astra and lands next to it on Arena.
  • Use Astra for the hardest science, math and long-running computer-use tasks, where OpenAI's own tables still show it ahead, and when the Pro tier in ChatGPT matters.
  • Use Luna for classification, extraction, routing and other high-volume calls, especially with none effort or the Decisions API. At $0.50 per million output tokens it is 20 times cheaper than Sol on output.
  • Check the 272K line. If prompts often cross it, the 2x/1.5x multiplier applies to the whole request on every GPT-6 model.

Compare before you choose: pricing changes and rankings move daily. Check the current model rankings on AI Rank before you commit.

Or 📱 get the AI Rank app to follow them on your phone.

Sources

Prepared with automated editorial assistance and checked on October 22, 2026 against OpenAI's launch posts, model pages, pricing page and API changelog, plus dated AI Rank Arena and OpenRouter data. We did not run our own benchmarks for this article.