Back to Blog
AI News

Claude Opus 5.5: Pricing, Benchmarks and What's New

Claude Opus 5.5 launched September 22, 2026 at $4 / $20 per million tokens. See benchmarks, specs, the 5.5 lineup and how it compares with GPT-6 Astra.

8 min read
By OndaLabs Team

Claude Opus 5.5 is Anthropic's flagship model in the Claude 5.5 family, released September 22, 2026. As of October 17, 2026, it costs $4 per million input tokens and $20 per million output tokens, has a 1M-token context window, and ranks first on the Artificial Analysis Intelligence Index.

This guide covers the release, pricing, specs, benchmarks, comparisons with GPT-6 Astra and Gemini 4 Argon, and upgrade notes. Vendor figures are labelled as vendor-reported.

Claude Opus 5.5 release date and the 5.5 family

Anthropic released Opus 5.5 on September 22, 2026, calling it the first model in the Claude 5.5 family. The rest of the family followed:

  • September 22, 2026: Claude Opus 5.5
  • September 28, 2026: Claude Sonnet 5.5
  • October 7, 2026: Claude Haiku 5.5

Sources: Anthropic, Claude Opus 5.5, Anthropic, Claude Sonnet 5.5, and Anthropic, Claude Haiku 5.5.

Anthropic says Opus 5.5 performs at the level of Claude Fable 5.1 on most work. It also says the model costs "40% less to run than Opus 5" and generates output more than 30% faster. Both are vendor claims. Per-token prices tell a different story: Opus 5.5 is 20% cheaper than Opus 5 per token, so the 40% figure is about the cost to run a task, not the price on the page.

Claude Opus 5.5 pricing

Standard prices are per million tokens. The table compares the current Claude lineup with OpenAI and Google models that are often considered alongside it.

| Model | Input | Output | Context window | |---|---|---|---| | Claude Fable 5.1 | $10 | $50 | 1M tokens | | Claude Opus 5.5 | $4 | $20 | 1M tokens | | Claude Opus 5 (previous) | $5 | $25 | 1M tokens | | Claude Sonnet 5.5 | $2 | $10 | 1M tokens | | Claude Haiku 5.5 (up to 100K prompt) | $0.10 | $0.50 | 1M tokens | | GPT-6 Astra (short context) | $10 | $50 | 1.05M tokens | | Gemini 4 Argon | $2.00 | $10.00 | 1,048,576 tokens |

Prices from Anthropic's models overview and OpenAI's and Google's pricing pages, read October 17, 2026. Gemini 4 Argon's price rises to $4.00 input and $20.00 output from December 1, 2026, according to Google.

Opus 5.5 also has these rates:

  • Cache reads cost $0.20 per million tokens, which is 5% of the base input price.
  • Cache writes cost $5 per million tokens for the 5-minute cache and $8 for the 1-hour cache.
  • The Batch API is 50% off the standard price.
  • Fast mode, which Anthropic says runs up to 2.5x faster in Claude Code and on the Claude Platform, costs $8 input and $40 output per million tokens.

Source: Claude models overview.

Claude Opus 5.5 specs

| Specification | Claude Opus 5.5 | |---|---| | Context window | 1M tokens | | Max output | 128K tokens (300K on the Message Batches API with the output-300k-2026-03-24 beta header) | | Inputs | Text and images | | Outputs | Text | | Thinking | Adaptive, always on (cannot be disabled) | | Default effort | medium | | Reliable knowledge cutoff | June 2026 | | Claude API model ID | claude-opus-5-5 | | Bedrock model ID | anthropic.claude-opus-5-5 |

Anthropic says the model is available on all major clouds, including Amazon Web Services, Google Cloud and Microsoft Azure, and supports zero data retention. The earliest retirement date is September 22, 2027.

Source: Claude models overview.

Claude Opus 5.5 benchmarks

Anthropic's launch table compares Opus 5.5 with its predecessor, Fable 5.1, and two OpenAI models. These are Anthropic's figures, read from its announcement, and they are vendor-reported.

| Benchmark (vendor-reported) | Opus 5.5 | Fable 5.1 | Opus 5 | GPT-6 Astra | GPT-5.6 Sol | |---|---|---|---|---|---| | Terminal-Bench 4.0 | 66.4% | 55.8% | 52.3% | 57.9% | 37.3% | | FrontierCode v1.1 (Main) | 54.4% | 50.3% | 48.0% | 53.3% | 47.5% | | CursorBench 4.0 | 57.8% | 51.8% | 46.6% | not reported | 41.7% | | GDPval-AA v2.1 (Elo) | 1846 | 1735 | 1708 | 1542 | 1588 | | AutomationBench | 40.0% | 31.4% | 26.9% | 41.4% | 28.8% | | Humanity's Last Exam with tools | 67.7% | 65.6% | 63.6% | 57.2% | not reported | | Terminal-Bench-Science 0.1 | 58.7% | 52.6% | 29.0% | 64.6% | 22.4% |

Anthropic includes several caveats with its table:

  • Benchmark margins are a less reliable guide to real-world differences.
  • Opus 5.5 was tested with production safeguards turned on, which Anthropic says likely lowered some scores.
  • AutomationBench counted safeguard interventions as failures.
  • The GPT-6 Astra Terminal-Bench figures are "as reported by OpenAI."

Google's launch table for Gemini 4 Argon also includes Opus 5.5. In that table, Opus 5.5 scores 67.0% on the Vals Index against 68.9% for Argon. It also leads on Terminal-bench 4.0 (66.4%) and PostTrainBench (49.3%), and it scores 62.3% on FrontierSWE v2 and 74.2% on DeepSWE v1.1. Source: Google DeepMind's Gemini models page.

How Opus 5.5 ranks on the Artificial Analysis index

Artificial Analysis (AA) is an independent benchmark site. On October 17, 2026, it gave Claude Opus 5.5 at max effort (with the default fallback configuration) an Intelligence Index score of 58, ranked first of 217 models. Its output speed is 97.5 tokens per second, and its blended price is $2.94 per million tokens. One run of the index costs $5.98 at max effort. AA counts 260 million output tokens used on that run.

Effort matters a lot. AA's leaderboard shows how the score and cost change as effort goes down:

| Model (effort) | Intelligence Index | Cost per index task | |---|---|---| | Claude Opus 5.5 (max) | 58 | $5.98 | | Claude Opus 5.5 (xhigh) | 56 | $3.46 | | Claude Opus 5.5 (high) | 54 | $1.82 | | Claude Opus 5.5 (medium) | 51 | $1.34 | | Claude Sonnet 5.5 (max) | 56 | $5.46 | | Claude Fable 5.1 (max) | 53 | $7.63 | | GPT-6 Astra (max) | 53 | $3.26 | | Gemini 4 Argon (high) | 53 | $1.99 |

Source: Artificial Analysis leaderboard, and the AA's page for Claude Opus 5.5.

The effort setting is a real choice. Opus 5.5 at high effort scores 54 at $1.82 per index run, and at medium it scores 51 at $1.34. Anthropic's default effort for Opus 5.5 is medium.

Opus 5.5 vs GPT-6 Astra vs Gemini 4 Argon

Opus 5.5 leads the AA index at max effort. GPT-6 Astra is the most expensive of the three per output token, and Gemini 4 Argon has the lowest cost per index run in this table. The three models are listed below at their standard short-context API prices.

| Model | Input per 1M | Output per 1M | Context | AA Intelligence Index | |---|---|---|---|---| | Claude Opus 5.5 | $4 | $20 | 1M | 58 (max) | | GPT-6 Astra | $10 | $50 | 1.05M | 53 (max) | | Gemini 4 Argon | $2.00 | $10.00 | 1,048,576 | 53 (high) |

Pick the model on the task, not the headline. Astra leads Anthropic's table on AutomationBench and Terminal-Bench-Science, and Argon leads Google's Vals Index. For a wider comparison, see The best AI models right now: October 2026 roundup and GPT-6 Astra: features, pricing and how it compares.

Which Claude 5.5 model should you pick?

Anthropic's own guidance on the overview page is a useful starting point:

  • Opus 5.5 for most workloads.
  • Fable 5.1 for demanding reasoning and long-horizon agentic work, or when Opus 5.5 at higher effort is not good enough.
  • Sonnet 5.5 for a balance of speed and intelligence. Anthropic says its output is over 30% faster than Sonnet 5, at up to 30% lower cost per task, and it is priced at $2 and $10 per million tokens.
  • Haiku 5.5 for high-volume, latency-sensitive work. It is priced from $0.10 and $0.50, and rises to $0.50 and $2.50 for prompts above 100K tokens. It is the first Haiku with adjustable effort, from low to max, and uses an updated tokenizer.

Anthropic calls Haiku 5.5 "the cheapest, fastest, and most capable small model" it has released. That is Anthropic's wording and should be read as its claim.

Upgrading from Opus 5 to Opus 5.5

The overview page lists breaking changes you should check before switching:

  1. Thinking cannot be disabled on Opus 5.5.
  2. Forced tool use returns an error.
  3. Thinking blocks are tied to the model and to the conversation they came from.
  4. On the Claude API and Google Cloud, the earlier computer_20251124 tool is not accepted.
  5. Text between tool calls comes back inside thinking blocks with empty text at the default display setting.

Source: Claude models overview.

Where voice fits

Opus 5.5 returns text. If your app reads answers aloud, a text to speech model does the voice work. OndaLabs is an AI voice studio. Its text to speech includes 30 prebuilt voices, usable in every supported language, and 37 language pages. Onda Lite costs 1 credit per character on all plans, and Onda Flash costs 1.5 credits per character on paid plans. Speech to text costs 50 credits per minute, for turning recorded audio into text. Plan prices are on the pricing page.

Frequently asked questions

What is Claude Opus 5.5?

Claude Opus 5.5 is Anthropic's flagship model in the Claude 5.5 family, released September 22, 2026. It accepts text and images and returns text. It has a 1M-token context window and adaptive thinking that is always on.

When was Claude Opus 5.5 released?

Anthropic released Claude Opus 5.5 on September 22, 2026. Sonnet 5.5 followed on September 28, and Haiku 5.5 on October 7, 2026.

How much does Claude Opus 5.5 cost?

Opus 5.5 costs $4 per million input tokens and $20 per million output tokens at standard rates. Cache reads cost $0.20 per million tokens, and the Batch API is 50% off. Fast mode costs $8 input and $40 output per million tokens.

How is Opus 5.5 different from Opus 5 and Fable 5.1?

Opus 5.5 costs less per token than Opus 5, at $4 and $20 against $5 and $25. Anthropic says Opus 5.5 performs at the level of Fable 5.1 on most work, and Fable 5.1 is priced higher at $10 and $50. Thinking cannot be turned off on Opus 5.5, so check your code for that before you upgrade.

Is Claude Opus 5.5 better than GPT-6 Astra or Gemini 4 Argon?

On the Artificial Analysis Intelligence Index on October 17, 2026, Opus 5.5 at max effort scored 58, ahead of Astra and Argon at 53 each. Anthropic's table shows Opus 5.5 ahead of Astra on most rows, but not all. Vendor benchmarks differ, so test on your own tasks.

What is the context window and knowledge cutoff of Claude Opus 5.5?

Opus 5.5 has a 1M-token context window and up to 128K output tokens. Anthropic gives a reliable knowledge cutoff of June 2026. Its model ID in the Claude API is claude-opus-5-5.