Skip to content
Models / Anthropic

Claude Opus 5.5

Anthropic · Model release

Released

Catalog checked 2026-09-28

Input $/1M
$4.00
Output $/1M
$20.00
Context
1M
Max output
128K
Reasoning
Yes
Included in
21 plans
API priceUSD / 1M tokens
Published API rates in US dollars per million tokens
RateInputOutputCache readCache write
StandardInput$4.00Output$20.00Cache read$0.20Cache write$5.00
Reasoning tokens
Billed at the output token rate

Published rate checked 2026-09-23. Sources below.

Current published base rates. Context tiers, cache-write duration and other conditions may change the rate for a request. These are not a reconstructed historical invoice.

API token estimate

Explore a token scenario

Uncached text input and billed output at this exact model’s recorded Standard API list price. Totals across a scenario, without a claim about one-request capacity.

Standard API list price · USD / 1M tokens

Pricing checked Sep 23, 2026 · Rate record from Sep 22, 2026

Record dates describe catalog coverage; they do not establish provider price activation.

Pricing record: claude-opus-5-5-pricing

Capabilities & limitsProvider specifications
Context window
1,000,000
Maximum output
128,000
Input
text · image
Output
text
Knowledge cutoff
Jun 2026
Reasoning
Supported
Tool calling
Supported
Before you chooseChecked 2026-09-29
Thinking
Adaptive (always on) · default effort medium
API availability
Active (latest)
Retirement commitment
Not sooner than September 22, 2027
Extended output
Up to 300K output tokens on Message Batches with the output-300k-2026-03-24 beta header. Standard output limit remains 128K.
Prompt caching
Minimum eligible prompt: 512 tokens. Five-minute and one-hour cache writes have different prices.

Google DeepMind · Published Sep 30, 2026

Developer reported · Gemini 4 Argon launch

Show all 17 results
Methodology & sources

Google’s launch sheet combines evaluations it computed with results it republished from leaderboards and system cards. This is one reporting set; it does not establish identical harnesses across models. Argon generally uses the Gemini API at the highest thinking setting, with a single attempt and trial averaging on smaller benchmarks.

These scores are reported by Google DeepMind. Model developers identify authorship; results may originate with another evaluator. Exact benchmark versions remain distinct. StackReplay calculates no composite score.

DeepSWE v1.1 · 74.2%

Long-horizon software engineering in real-world codebases.

Version: v1.1 · Reported benchmark score · percent · Higher is better · Task subset: Not separately specified

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
Leaderboard or system card cited by Google (republished result)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
Not reported
Tools
Not reported
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Argon uses Google’s mini-swe agent; Astra is reported from the leaderboard, Fable and Opus from system cards. Thinking levels follow the highest scoring level reported by Datacurve.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

FrontierSWE v2 · 62.3%

Agentic software engineering on coding tasks.

Version: v2 · Reported benchmark score · percent · Higher is better · Task subset: Not separately specified

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
Proximal public leaderboard (republished result)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
Not reported
Tools
Not reported
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Results republished from Proximal’s public leaderboard.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

Vibe Code Bench · 90.3%

Code generation for prompted software-building tasks.

Version: Not reported · Reported benchmark score · percent · Higher is better · Task subset: Not separately specified

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
Vals AI public leaderboard (republished result)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
Not reported
Tools
Not reported
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Results republished from the Vals AI public leaderboard.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

Terminal-Bench 4.0 · 66.4%

Agentic task completion in a terminal environment.

Version: 4.0 · Reported benchmark score · percent · Higher is better · Task subset: Not separately specified

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
Terminal-Bench public leaderboard (republished result)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
Not reported
Tools
Not reported
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Argon is computed by Google; competitor results are republished from the official leaderboard, using the highest scoring thinking level reported by its authors.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

PostTrainBench v1.1 · 49.3%

Post-training of language models under a fixed compute budget.

Version: v1.1 · Benchmark-reported weighted score · percent · Higher is better · Task subset: Not separately specified

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
Google DeepMind (source-computed)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
OpenCode
Tools
Single NVIDIA H100, 10-hour budget
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Google computed every model with OpenCode and a 10-hour budget on one NVIDIA H100; the benchmark’s metric aggregates four base models and seven benchmarks.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

Vals Index · 67.0%

Economic knowledge work across finance, coding, legal and tax tasks.

Version: Not reported · Reported benchmark score · percent · Higher is better · Task subset: Not separately specified

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
Vals AI (republished result)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
Not reported
Tools
Not reported
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Vals AI results republished by Google.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

AutomationBench · 42.5%

End-to-end execution of business workflows.

Version: Not reported · Reported benchmark score · percent · Higher is better · Task subset: Not separately specified

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
Zapier public leaderboard (republished result)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
Not reported
Tools
Not reported
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Private task set; results republished from Zapier’s public leaderboard.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

Vals Finance Agent v2 · 58.6%

Multi-step financial research by an agent.

Version: v2 · Reported benchmark score · percent · Higher is better · Task subset: Not separately specified

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
Vals AI (republished result)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
Not reported
Tools
Not reported
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Vals AI results republished by Google.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

Harvey’s Legal Agent Benchmark · 3.8%

Legal research and drafting by an agent.

Version: Not reported · Reported benchmark score · percent · Higher is better · Task subset: Not separately specified

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
Vals AI (republished result)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
Not reported
Tools
Not reported
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Vals AI results republished by Google.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

Terminal-Bench Science 0.1 · 63.3%

Scientific and mathematical tasks in a terminal environment.

Version: 0.1 · Reported benchmark score · percent · Higher is better · Task subset: Not separately specified

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
Terminal-Bench Science public leaderboard (republished result)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
Not reported
Tools
Not reported
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Argon uses a sixfold verifier timeout to address verification timeouts; competitor results come from the official leaderboard.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

LABBench 2 · 73.1%

Scientific tasks with a terminal, bioinformatics tools and internet access.

Version: 2 · Reported benchmark score · percent · Higher is better · Task subset: Not separately specified

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
Google DeepMind (source-computed)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
Not reported
Tools
Linux terminal, bioinformatics tools, Python, R and internet
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Google computed every model with a Linux terminal, bioinformatics tools, Python, R and internet access.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

RiemannBench · 69.6%

Mathematical problem solving.

Version: Not reported · Reported benchmark score · percent · Higher is better · Task subset: Not separately specified

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
Surge public leaderboard (republished result)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
Not reported
Tools
Not reported
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Results republished from Surge’s public leaderboard.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

GraphWalks · up to 128k BFS F1 · 90.6%

Breadth-first graph traversal with contexts up to 128k tokens, scored by F1.

Version: Not reported · BFS F1 · percent · Higher is better · Task subset: · up to 128k BFS F1

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
Google DeepMind (source-computed)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
Not reported
Tools
Not reported
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Google computed every model on 650 problems with context lengths up to 128k tokens.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

GraphWalks · 256k–1M BFS F1 · 66.8%

Breadth-first graph traversal with 256k to 1M-token contexts, scored by F1.

Version: Not reported · BFS F1 · percent · Higher is better · Task subset: · 256k–1M BFS F1

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
Google DeepMind (source-computed)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
Not reported
Tools
Not reported
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Google computed every model on 200 problems with context lengths from 256k to 1M tokens.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

Chartography · 66.3%

Visual interpretation and analysis of charts.

Version: Not reported · Reported benchmark score · percent · Higher is better · Task subset: Not separately specified

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
Surge public leaderboard (republished result)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
Not reported
Tools
No tools
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Without tools; results republished from Surge’s public leaderboard.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

LVBench · 83.7%

Understanding information in long videos.

Version: Not reported · Reported benchmark score · percent · Higher is better · Task subset: Not separately specified

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
Google DeepMind (source-computed)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
Not reported
Tools
No tools
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Google computed every model without tools. Video sampling differs due to API limits: Argon 1 FPS, Astra 800 frames, Fable 300 frames, Opus 600 frames.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

CWE-bench v1 · 67.0%

Software vulnerability remediation.

Version: v1 · Reported benchmark score · percent · Higher is better · Task subset: Not separately specified

Google DeepMind · Published Sep 30, 2026 · Developer reported

Reporting source
Google DeepMind · Gemini 4 Argon launch
Evaluation origin
CWE-bench public leaderboard (republished result)
Effort
Maximum available reasoning; best available reported result when maximum-setting results are unavailable
Harness
Not reported
Tools
Not reported
Fallback
Not reported
Provider
Not reported
Checked
Sep 30, 2026

Pass@1 results republished from the official leaderboard. Its pass@4 tiebreak is not part of this sheet, so equal pass@1 scores remain ties.

No confidence interval reported in this comparison.

Original methodology ↗ · Original evidence ↗

OpenAI · Published Sep 29, 2026

Developer reported · GPT-6.1 Sol launch · Max effort

Knowledge work

GDP.pdf
26.2%
Methodology & sources

Launch chart results at Max reasoning effort. GPT evaluations ran in OpenAI research environments or its API. Competitor scores were taken from public reports. Exact chart values, versions, subsets and documented fallbacks are preserved.

These scores are reported by OpenAI. Model developers identify authorship; results may originate with another evaluator. Exact benchmark versions remain distinct. StackReplay calculates no composite score.

GDP.pdf · 26.2%

Answering professional questions about complex PDF documents.

Version: Not reported · Reported benchmark score · percent · Higher is better · Task subset: Not separately specified

OpenAI · Published Sep 29, 2026 · Developer reported

Reporting source
OpenAI · GPT-6.1 Sol launch · Max effort
Evaluation origin
Public competitor reports, as republished by OpenAI; original runner not identified (republished result)
Effort
Max
Harness
Not reported
Tools
Not reported
Fallback
Opus 5 / Opus 4.8 fallbacks
Provider
Not reported
Checked
Sep 30, 2026

Exact percentage converted from the numerical fraction embedded in the official launch chart. Harness, fallback and tool details are retained only where the publication specifies them.

Original methodology ↗ · Original evidence ↗

AutomationBench 1.0.6 · 42.47%

Completing multi-step business workflows with 47 available tools.

Version: 1.0.6 · Reported benchmark score · percent · Higher is better · Task subset: Not separately specified

OpenAI · Published Sep 29, 2026 · Developer reported

Reporting source
OpenAI · GPT-6.1 Sol launch · Max effort
Evaluation origin
Public competitor reports, as republished by OpenAI; original runner not identified (republished result)
Effort
Max
Harness
Not reported
Tools
47 tools across sales, marketing, operations, support, finance and HR
Fallback
Fallbacks used; releases not specified in this chart
Provider
Not reported
Checked
Sep 30, 2026

Exact percentage converted from the numerical fraction embedded in the official launch chart. Harness, fallback and tool details are retained only where the publication specifies them.

Original methodology ↗ · Original evidence ↗

Epoch AI · Archive checked Oct 4, 2026

Independent evaluation · Epoch AI Capabilities and benchmarking: GPQA Diamond mean accuracy

Methodology & sources

Epoch AI ran and reported this selected GPQA Diamond evaluation. Attribution: Epoch AI, 'Capabilities & benchmarking' (https://epoch.ai/benchmarks), original data (https://epoch.ai/data/benchmark_data.zip), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). StackReplay selected three Epoch-run records from the Diamond subset and converted accuracy fractions to percentages; no endorsement is implied. Current public methodology v1.0.6 is contextual only; run-specific suite and setup details are unknown.

These scores are reported by Epoch AI. Model developers identify authorship; results may originate with another evaluator. Exact benchmark versions remain distinct. StackReplay calculates no composite score.

GPQA Diamond Epoch runs; suite revision unreported · 90.5934343434343400%

Graduate-level scientific question answering administered by Epoch AI. The published Diamond subset contains 198 questions; actual scored counts and run-specific suite revisions are unreported.

Version: Not reported · Epoch-published mean accuracy · percent · Higher is better · Task subset: Diamond

Epoch AI · Archive checked Oct 4, 2026 · Independent evaluation

Reporting source
Epoch AI · Epoch AI Capabilities and benchmarking: GPQA Diamond mean accuracy
Evaluation origin
Epoch AI administered the evaluation (source-computed)
Effort
Max (Epoch source configuration label)
Harness
Inspect (general methodology context; run version unreported)
Tools
Not reported
Fallback
Not reported
Provider
Not reported
Checked
Oct 4, 2026

Source mean_score was the accuracy fraction 0.9059343434343434. Exact conversion from the source decimal string is 90.5934343434343400%. The raw stderr is 0.019434397190192135. Epoch published run ID 6gmbahPSxHf9xVhAYqpnhi and evaluation start timestamp 2026-09-22T18:27:39.000Z. The source configuration label is max. Completion and original publication dates are unknown. Source artifact: gpqa_diamond.csv in the checked archive. ZIP SHA256 86e6667353c6c8c396feaa84eabc5d9ce160ee6fa22b2cd5abb72e26614e788c; complete archive-member SHA256 c5fed6f394d650dada5c5d14c325a0ccbd61b742b555b95ca971d5a1fed3b4d1. The download URL is mutable; these hashes identify the checked bytes.

Suite revision, scored count, repetitions, total attempts, scaffold revision, prompt, temperature, reasoning budget, output/time/cost caps, endpoint, tools, seed, error handling, and completion date are unreported. The source configuration label does not establish equal effort across providers.

Original methodology ↗ · Original evidence ↗

Where you can use it1 API route · 21 subscriptions
Anthropic APIDirect API
Claude Pro
$20 / month ↗

Monthly or annual. Canceling keeps access until the paid period ends.

Claude Max 5x
$100 / month ↗

Monthly only; web upgrades are prorated. Canceling keeps access until the paid period ends.

Claude Max 20x
$200 / month ↗

Monthly only; web upgrades are prorated. Canceling keeps access until the paid period ends.

Command Code Max 10×
$100 / month ↗

Monthly cycle; purchased credits carry over without expiry.

Command Code Max 20×
$200 / month ↗

Monthly cycle; purchased credits carry over without expiry.

Cursor Pro
$20 / month ↗

Monthly pool reset follows billing cycle; on-demand usage billed in arrears.

Cursor Teams Standard
$40 per paid user / month ↗

Per paid user; monthly pools reset each billing cycle. A yearly toggle exists, but commitment terms are unpublished.

Absolute pool sizes, annual commitment, minimum seats, trial availability and country eligibility are not published.

Cursor Pro+
$60 / month ↗

Monthly pool reset follows billing cycle; on-demand usage billed in arrears.

Cursor Teams Premium
$120 per paid user / month ↗

Per paid user; monthly pools reset each billing cycle. A yearly toggle exists, but commitment terms are unpublished.

Premium's 5x claim is relative to an unpublished Standard amount. Absolute pool sizes, annual commitment, minimum seats, trial availability and country eligibility are not published.

Cursor Ultra
$200 / month ↗

Monthly pool reset follows billing cycle; on-demand usage billed in arrears.

Copilot Business
$19 / month ↗

Credits reset on the first of the month; unused included credits do not carry over.

Copilot Enterprise
$39 / month ↗

Credits reset on the first of the month; unused included credits do not carry over.

Copilot Pro+
$39 / month ↗

Credits reset on the first of the month; unused included credits do not carry over.

Copilot Max
$100 / month ↗

Credits reset on the first of the month; unused included credits do not carry over.

Google AI Pro
$19.99 / month ↗

Monthly plan; each product has separate usage limits and credit rules.

Google-login Gemini CLI and IDE access for Free, Google AI Pro and Google AI Ultra ceased June 18, 2026. Licensed Code Assist Standard and Enterprise are unaffected.

Google AI Ultra 5x
$99.99 / month ↗

Monthly plan; each product has separate usage limits and credit rules.

Google-login Gemini CLI and IDE access for Free, Google AI Pro and Google AI Ultra ceased June 18, 2026. Licensed Code Assist Standard and Enterprise are unaffected.

Google AI Ultra 20x
$199.99 / month ↗

Monthly plan; each product has separate usage limits and credit rules.

Google-login Gemini CLI and IDE access for Free, Google AI Pro and Google AI Ultra ceased June 18, 2026. Licensed Code Assist Standard and Enterprise are unaffected.

Kiro Pro
$20 per user / month ↗

Monthly per user; included credits reset monthly and do not roll over; no annual option is published.

Kiro Pro+
$40 per user / month ↗

Monthly per user; included credits reset monthly and do not roll over; no annual option is published.

Kiro Pro Max
$100 per user / month ↗

Monthly per user; included credits reset monthly and do not roll over; no annual option is published.

Kiro Power
$200 per user / month ↗

Monthly per user; included credits reset monthly and do not roll over; no annual option is published.

Related releasesOpus
Claude Opus 4.5Legacy ↗Claude Opus 4.6Legacy ↗Claude Opus 4.7Legacy ↗Claude Opus 4.8Legacy ↗Claude Opus 4.8 (fast mode) (preview)Legacy ↗Claude Opus 5Legacy ↗
Pricing, assumptions & evidence

Published base rate

Input $4.00 · Output $20.00 · Cache read $0.20 · Cache write $5.00

Missing token-category prices are not zero. Workload pricing applies exact recorded categories and admitted routes. No benchmark score or quality ranking is inferred from prices.

Aliases, routes and identity

Catalog ID claude-opus-5-5

Model release (kind: release)

Developer: Anthropic

claude-opus-5-5

provider_id