RankingClaude Opus 5.5

Claude Opus 5.5

Data updated 31 Oct 2026

51 published benchmark measures · 1 benchmark families contribute across 1 task areas. 1 capability estimates available in the full profile. See all results ↓ · Compare published benchmarks →

Model evidence summary

This profile combines published settings. It is not a runnable configuration or a leaderboard rank. Compare measured configurations →

Performance profile

Capabilities

Adjusted comparison score · 0–100

Filled points are supported; hollow points are preliminary. Lines stop at unknown capabilities. Exact values and sources follow below.

AgenticNo comparable evidenceUnknown

0 core families; 0 direct opponents across 0 labs. No admitted core comparisons

Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.

Refit without one family: No supported estimate; 13/13 scenarios unsupported. Without one publisher: No supported estimate; 15/15 unsupported. Smoothing check: No supported estimate; 3/3 unsupported. These are sensitivity checks, not confidence intervals.

Audit: family and publisher holdouts improved on neutral predictions within evaluated coverage; source sensitivity still requires inspection. Validation audit.

Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.

Hard reasoning1 families · 1 with independent evidence · Preliminary83.8

1 core families; 59 direct opponents across 10 labs. Needs broader benchmark and opponent support

Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.

Refit without one family: No supported estimate; 6/6 scenarios unsupported. Without one publisher: No supported estimate; 16/16 unsupported. Smoothing check: No supported estimate; 3/3 unsupported. These are sensitivity checks, not confidence intervals.

Audit: family and publisher holdouts improved on neutral predictions within evaluated coverage. Validation audit.

Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.

ARC-AGI: 96.6 observed win share
arcprize.org · Source 1

CodingNo comparable evidenceUnknown

0 core families; 0 direct opponents across 0 labs. No admitted core comparisons

Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.

Refit without one family: No supported estimate; 6/6 scenarios unsupported. Without one publisher: No supported estimate; 19/19 unsupported. Smoothing check: No supported estimate; 3/3 unsupported. These are sensitivity checks, not confidence intervals.

Audit: family and publisher holdouts improved on neutral predictions within evaluated coverage. Validation audit.

Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.

Human prefNo comparable evidenceUnknown

0 core families; 0 direct opponents across 0 labs. No admitted core comparisons

Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.

Refit without one family: No supported estimate; 1/1 scenarios unsupported. Without one publisher: No supported estimate; 1/1 unsupported. Smoothing check: No supported estimate; 3/3 unsupported. These are sensitivity checks, not confidence intervals.

Audit: cross-family and cross-publisher predictions could not be evaluated from the available comparison network. Validation audit.

Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.

KnowledgeNo comparable evidenceUnknown

0 core families; 0 direct opponents across 0 labs. No admitted core comparisons

Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.

Refit without one family: No supported estimate; 4/4 scenarios unsupported. Without one publisher: No supported estimate; 5/5 unsupported. Smoothing check: No supported estimate; 3/3 unsupported. These are sensitivity checks, not confidence intervals.

Audit: almost no publisher-held-out comparisons remained estimable. Cross-source validity is unresolved. Validation audit.

Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.

MultimodalNo comparable evidenceUnknown

0 core families; 0 direct opponents across 0 labs. No admitted core comparisons

Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.

Refit without one family: No supported estimate; 8/8 scenarios unsupported. Without one publisher: No supported estimate; 8/8 unsupported. Smoothing check: No supported estimate; 3/3 unsupported. These are sensitivity checks, not confidence intervals.

Audit: limited publisher-held-out coverage. Cross-source validity is unresolved. Validation audit.

Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.

Long contextNo comparable evidenceUnknown

0 core families; 0 direct opponents across 0 labs. No admitted core comparisons

Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.

Refit without one family: No supported estimate; 5/5 scenarios unsupported. Without one publisher: No supported estimate; 5/5 unsupported. Smoothing check: No supported estimate; 3/3 unsupported. These are sensitivity checks, not confidence intervals.

Audit: family and publisher holdouts underperformed neutral predictions. Treat this as an exploratory estimate. Validation audit.

Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.

Scores estimate outcomes against a shared reference panel; they are not accuracy percentages. Sparse or disconnected evidence cannot qualify an overall profile. Open a capability to inspect its evidence.

Reported effort · Mixed settings

Settings reported in this model's published benchmark results, including results outside the aggregate. Effort names are provider-specific. These are not API defaults or equal compute budgets.

Auto thinking: 2 observations
High: 4 observations
Low: 2 observations
Max: 34 observations
Medium: 5 observations
XHigh: 6 observations
Not specified: 37 observations

Mixed settings means multiple settings occur in the evidence. Best across efforts means the source selected its best reported result across settings; it does not mean Max. Unspecified settings stay unknown. This model-summary chart combines reported settings. The leaderboard keeps identified configurations separate and excludes unknown effort.

Inspect each result and its source ↓ · Download effort evidence

Score contributions and missing evidence

1 contributing families across 1 capabilities. Fixed reference panels do not change when the catalog expands.

Capability is fitted jointly across families. Capability estimates below describe different task areas; their weighted sum is not the Capability score.

Results without reviewed compatibility or a reference match remain in the raw evidence below. Coverage counts only contributing results.

Model information & shareable badge

Lab: Anthropic
Catalog status: active
Availability: Public provider catalog; account and region restrictions may apply
Family: Claude Opus
Released: 2026-09-22
Context: 1,000,000 tokens
API list price: $4 input / $20 output per million tokens
License: proprietary
Model card: https://platform.claude.com/docs/en/models/opus-5-5/overview
Default Capability family coverage: /badge/claude-opus-5-5.svg

Benchmark scores & sources

Original results, evaluation harnesses, and evidence behind this model.

Benchmark	Bucket	Score	Harness	Evidence	Source-recorded date
GraphWalks · up to 128K BFS F1	Supporting evidence	90.6% Reported settings & source Comparator maximum available thinking/reasoning when reported, otherwise best available result; exact setting not identified by this grid. All models self-computed. 650 items with context length up to 128K tokens. BFS F1, not pass rate. Provider-published launch claim, not independent measurement by Google for every row. All models self-computed. 650 items with context length up to 128K tokens. BFS F1, not pass rate. Methodology https://deepmind.google/models/evals-methodology/gemini-4-argon, SHA256 ff1df6bdeddc4c0f48840c09a8a7813b10ae7ad9e053892da68dabf64f09bff7. Preserve source rounding. Methodology says capabilities as of September 2026 and results as of October 2026; review date is source retrieval, not evaluation date. Comparison limit: Provider-published claim. Independent source results are admitted separately with exact checkpoint, effort and compatible protocol; this grid does not establish a matched comparison. Google Gemini 4 Argon launch and evaluation methodology · Launch grid row 13: GraphWalks up to 128K BFS F1 / Claude Opus 5.5; methodology Additional Details · reviewed 2026-09-30	Published configuration Effort: Not specified	lab self-report Provider-published claim. Independent source results are admitted separately with exact checkpoint, effort and compatible protocol; this grid does not establish a matched comparison.	Reviewed 2026-09-30
Harvey's Legal Agent Benchmark · 1	Supporting evidence	3.8% Reported settings & source Comparator maximum available thinking/reasoning when reported, otherwise best available result; exact setting not identified by this grid. Results sourced from Vals AI. Provider-published launch claim, not independent measurement by Google for every row. Results sourced from Vals AI. Methodology https://deepmind.google/models/evals-methodology/gemini-4-argon, SHA256 ff1df6bdeddc4c0f48840c09a8a7813b10ae7ad9e053892da68dabf64f09bff7. Preserve source rounding. Methodology says capabilities as of September 2026 and results as of October 2026; review date is source retrieval, not evaluation date. Comparison limit: Provider-published claim. Independent source results are admitted separately with exact checkpoint, effort and compatible protocol; this grid does not establish a matched comparison. Google Gemini 4 Argon launch and evaluation methodology · Launch grid row 4: Harvey's Legal Agent Benchmark 1 / Claude Opus 5.5; methodology Additional Details · reviewed 2026-09-30	Published configuration Effort: Not specified	lab self-report Provider-published claim. Independent source results are admitted separately with exact checkpoint, effort and compatible protocol; this grid does not establish a matched comparison.	Reviewed 2026-09-30
HealthBench	Supporting evidence	68.1% Reported settings & source Raw rubric score; adaptive thinking at max effort; five trials; no tools or custom system prompt; Opus 4.8 grader; safety classifiers with refusal fallback to Opus 5. Provider-published lab self-report from the Claude Opus 5.5 system card (SHA256 7311c9c6bbb16d012f1c12c7418b05949fcf7ae3e30d2c40f22050074b2a7378). Not an independent board. Comparison limit: Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison. Claude Opus 5.5 System Card · section 8.15.1 · reviewed 2026-09-22	Published configuration Effort: Max	lab self-report Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison.	Reviewed 2026-09-22
HealthBench	Supporting evidence	60.6% Reported settings & source Length-adjusted score using the GPT-5.5 system-card method; otherwise the raw HealthBench configuration: adaptive max, five trials, no tools, Opus 4.8 grader, safety fallback to Opus 5. Provider-published lab self-report from the Claude Opus 5.5 system card (SHA256 7311c9c6bbb16d012f1c12c7418b05949fcf7ae3e30d2c40f22050074b2a7378). Not an independent board. Comparison limit: Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison. Claude Opus 5.5 System Card · section 8.15.1 · reviewed 2026-09-22	Published configuration Effort: Max	lab self-report Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison.	Reviewed 2026-09-22
HealthBench Professional	Supporting evidence	77.1% Reported settings & source Raw rubric score; adaptive thinking at max effort; five trials; no tools or custom system prompt; Opus 4.8 grader; safety classifiers with refusal fallback to Opus 5. Provider-published lab self-report from the Claude Opus 5.5 system card (SHA256 7311c9c6bbb16d012f1c12c7418b05949fcf7ae3e30d2c40f22050074b2a7378). Not an independent board. Comparison limit: Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison. Claude Opus 5.5 System Card · section 8.15.2 · reviewed 2026-09-22	Published configuration Effort: Max	lab self-report Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison.	Reviewed 2026-09-22
HealthBench Professional	Supporting evidence	65.6% Reported settings & source Length-adjusted score using the HealthBench Professional paper method; adaptive max; five trials; no tools; Opus 4.8 grader; safety fallback to Opus 5. Provider-published lab self-report from the Claude Opus 5.5 system card (SHA256 7311c9c6bbb16d012f1c12c7418b05949fcf7ae3e30d2c40f22050074b2a7378). Not an independent board. Table 8.1.A's 65.6 is this length-adjusted score, not the raw 77.1%. Comparison limit: Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison. Claude Opus 5.5 System Card · Table 8.1.A; section 8.15.2 · reviewed 2026-09-22	Published configuration Effort: Max	lab self-report Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison.	Reviewed 2026-09-22
Humanity’s Last Exam	Hard reasoning	64.4% Reported settings & source Full 2,500 questions; no tools; section 8.11.1 sets thinking to auto, a 1M total token cap, no compaction, and an Opus 4.6 grader. Provider-published lab self-report from the Claude Opus 5.5 system card (SHA256 7311c9c6bbb16d012f1c12c7418b05949fcf7ae3e30d2c40f22050074b2a7378). Not an independent board. Table 8.1.A's default note says adaptive max effort unless otherwise noted. Section 8.11.1 states thinking was set to auto for these runs. Comparison limit: Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison. Claude Opus 5.5 System Card · Table 8.1.A; section 8.11.1 · reviewed 2026-09-22	Published configuration Effort: Auto thinking	lab self-report Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison.	Reviewed 2026-09-22
Humanity’s Last Exam	Hard reasoning	67.7% Reported settings & source Full 2,500 questions; web search, web fetch, programmatic tool calling, and code execution; thinking set to auto; 1M total token cap; no compaction; Opus 4.6 grader; HLE source blocklist and contamination review. Provider-published lab self-report from the Claude Opus 5.5 system card (SHA256 7311c9c6bbb16d012f1c12c7418b05949fcf7ae3e30d2c40f22050074b2a7378). Not an independent board. Launch grid shows 67.7% with tools. Comparison limit: Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison. Claude Opus 5.5 System Card · Table 8.1.A; section 8.11.1 · reviewed 2026-09-22	Published configuration Effort: Auto thinking	lab self-report Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison.	Reviewed 2026-09-22
Humanity's Sixth Sense	Supporting evidence	44.6%	HSS · xhigh Effort: xhigh Reported settings Full 522-task pass@1, task mean of three attempts; every rubric criterion must pass. 95% task-cluster bootstrap CI. Claude-Opus-5 judge. Scale AI + Elorian. Unversioned board; source hashes identify captures. Display only. Excluded from weighted Overall. Source-recorded date is the capture date, not a proven evaluation date.	official board Supporting evidence outside the reviewed capability core	2026-10-31
LABBench · 2	Supporting evidence	73.1% Reported settings & source Comparator maximum available thinking/reasoning when reported, otherwise best available result; exact setting not identified by this grid. All models self-computed with Linux terminal, pre-installed bioinformatics tools, Python, R and internet access. Provider-published launch claim, not independent measurement by Google for every row. All models self-computed with Linux terminal, pre-installed bioinformatics tools, Python, R and internet access. Methodology https://deepmind.google/models/evals-methodology/gemini-4-argon, SHA256 ff1df6bdeddc4c0f48840c09a8a7813b10ae7ad9e053892da68dabf64f09bff7. Preserve source rounding. Methodology says capabilities as of September 2026 and results as of October 2026; review date is source retrieval, not evaluation date. Comparison limit: Provider-published claim. Independent source results are admitted separately with exact checkpoint, effort and compatible protocol; this grid does not establish a matched comparison. Google Gemini 4 Argon launch and evaluation methodology · Launch grid row 11: LABBench 2 / Claude Opus 5.5; methodology Additional Details · reviewed 2026-09-30	Published configuration Effort: Not specified	lab self-report Provider-published claim. Independent source results are admitted separately with exact checkpoint, effort and compatible protocol; this grid does not establish a matched comparison.	Reviewed 2026-09-30
Legal Agent Benchmark · 120-task held-out subset	Supporting evidence	8.3% Reported settings & source All-pass rate on Harvey's held-out 120 tasks; max effort. Artificial Analysis harness, as in the previous system card. Provider-published lab self-report from the Claude Opus 5.5 system card (SHA256 7311c9c6bbb16d012f1c12c7418b05949fcf7ae3e30d2c40f22050074b2a7378). Not an independent board. Comparison limit: Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison. Claude Opus 5.5 System Card · section 8.14.2 · reviewed 2026-09-22	Published configuration Effort: Max	lab self-report Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison.	Reviewed 2026-09-22
Legal Agent Benchmark · 120-task held-out subset	Supporting evidence	91.2% Reported settings & source Mean criterion-pass rate on Harvey's held-out 120 tasks; max effort. Artificial Analysis harness, as in the previous system card. Provider-published lab self-report from the Claude Opus 5.5 system card (SHA256 7311c9c6bbb16d012f1c12c7418b05949fcf7ae3e30d2c40f22050074b2a7378). Not an independent board. Comparison limit: Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison. Claude Opus 5.5 System Card · section 8.14.2 · reviewed 2026-09-22	Published configuration Effort: Max	lab self-report Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison.	Reviewed 2026-09-22
LMArena Agent Arena	Supporting evidence	20.94	LMArena Agent Arena · max Effort: Not specified	official board Supporting evidence outside the reviewed capability core	2026-10-28
LMArena Agent Arena	Supporting evidence	14.33	LMArena Agent Arena · high Effort: Not specified	official board Supporting evidence outside the reviewed capability core	2026-10-08
LVBench	Multimodal	83.7% Reported settings & source Comparator maximum available thinking/reasoning when reported, otherwise best available result; exact setting not identified by this grid. All models self-computed without tools. Gemini 1 FPS; Astra 800 frames, Fable 300 frames and Opus 600 frames due to API limits. Unequal frame budgets are excluded from matched comparison. Provider-published launch claim, not independent measurement by Google for every row. All models self-computed without tools. Gemini 1 FPS; Astra 800 frames, Fable 300 frames and Opus 600 frames due to API limits. Unequal frame budgets are excluded from matched comparison. Methodology https://deepmind.google/models/evals-methodology/gemini-4-argon, SHA256 ff1df6bdeddc4c0f48840c09a8a7813b10ae7ad9e053892da68dabf64f09bff7. Preserve source rounding. Methodology says capabilities as of September 2026 and results as of October 2026; review date is source retrieval, not evaluation date. Comparison limit: Provider-published claim. Independent source results are admitted separately with exact checkpoint, effort and compatible protocol; this grid does not establish a matched comparison. Google Gemini 4 Argon launch and evaluation methodology · Launch grid row 18: LVBench / Claude Opus 5.5; methodology Additional Details · reviewed 2026-09-30	Published configuration Effort: Not specified	lab self-report Provider-published claim. Independent source results are admitted separately with exact checkpoint, effort and compatible protocol; this grid does not establish a matched comparison.	Reviewed 2026-09-30
MILU	Knowledge	93.1% Reported settings & source Average accuracy across 11 languages; adaptive thinking at max effort; five trials; no tools or custom system prompt. Provider-published lab self-report from the Claude Opus 5.5 system card (SHA256 7311c9c6bbb16d012f1c12c7418b05949fcf7ae3e30d2c40f22050074b2a7378). Not an independent board. Comparison limit: Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison. Claude Opus 5.5 System Card · section 8.16.2; Figure 8.16.2.A · reviewed 2026-09-22	Published configuration Effort: Max	lab self-report Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison.	Reviewed 2026-09-22
OfficeQA	Agentic	78.9% Reported settings & source Agentic extracted-text Treasury Bulletin corpus with code execution; max effort; mean of five runs. Provider-published lab self-report from the Claude Opus 5.5 system card (SHA256 7311c9c6bbb16d012f1c12c7418b05949fcf7ae3e30d2c40f22050074b2a7378). Not an independent board. Comparison limit: Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison. Claude Opus 5.5 System Card · section 8.14.1 · reviewed 2026-09-22	Published configuration Effort: Max	lab self-report Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison.	Reviewed 2026-09-22
OfficeQA Pro	Agentic	67.7% Reported settings & source Harder 133-question subset; agentic extracted-text Treasury Bulletin corpus with code execution; max effort; mean of five runs. Provider-published lab self-report from the Claude Opus 5.5 system card (SHA256 7311c9c6bbb16d012f1c12c7418b05949fcf7ae3e30d2c40f22050074b2a7378). Not an independent board. Comparison limit: Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison. Claude Opus 5.5 System Card · section 8.14.1 · reviewed 2026-09-22	Published configuration Effort: Max	lab self-report Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison.	Reviewed 2026-09-22
OSWorld · 2.0 September 10, 2026 task release	Agentic	81.8% Reported settings & source Partial score, pass@1; 108 tasks; five runs; 1080p; 500 action steps; max reasoning effort; Opus 4.8 grader where required. September 10, 2026 task files and server-side context compaction after 100k tokens. Provider-published lab self-report from the Claude Opus 5.5 system card (SHA256 7311c9c6bbb16d012f1c12c7418b05949fcf7ae3e30d2c40f22050074b2a7378). Not an independent board. Launch grid shows 81.8% partial. Section 8.13.3 says these settings supersede the Fable 5.1 card's OSWorld 2.0 figures. Comparison limit: Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison. Claude Opus 5.5 System Card · Table 8.1.A; section 8.13.3 · reviewed 2026-09-22	Published configuration Effort: Max	lab self-report Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison.	Reviewed 2026-09-22
OSWorld · 2.0 September 10, 2026 task release	Agentic	48.7% Reported settings & source Strict pass rate, pass@1; 108 tasks; five runs; 1080p; 500 action steps; max reasoning effort; Opus 4.8 grader where required. September 10, 2026 task files and server-side context compaction after 100k tokens. Provider-published lab self-report from the Claude Opus 5.5 system card (SHA256 7311c9c6bbb16d012f1c12c7418b05949fcf7ae3e30d2c40f22050074b2a7378). Not an independent board. Table 8.1.A prints partial/strict as 81.8/48.7. This row is the strict pass rate. Comparison limit: Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison. Claude Opus 5.5 System Card · Table 8.1.A; section 8.13.3 · reviewed 2026-09-22	Published configuration Effort: Max	lab self-report Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison.	Reviewed 2026-09-22
PostTrainBench · 1.1	Supporting evidence	49.3% Reported settings & source Comparator maximum available thinking/reasoning when reported, otherwise best available result; exact setting not identified by this grid. All models self-computed; OpenCode harness; 10-hour budget on one NVIDIA H100 GPU. Weighted aggregate across four base models and seven benchmarks. Provider-published launch claim, not independent measurement by Google for every row. All models self-computed; OpenCode harness; 10-hour budget on one NVIDIA H100 GPU. Weighted aggregate across four base models and seven benchmarks. Methodology https://deepmind.google/models/evals-methodology/gemini-4-argon, SHA256 ff1df6bdeddc4c0f48840c09a8a7813b10ae7ad9e053892da68dabf64f09bff7. Preserve source rounding. Methodology says capabilities as of September 2026 and results as of October 2026; review date is source retrieval, not evaluation date. Comparison limit: Provider-published claim. Independent source results are admitted separately with exact checkpoint, effort and compatible protocol; this grid does not establish a matched comparison. Google Gemini 4 Argon launch and evaluation methodology · Launch grid row 9: PostTrainBench 1.1 / Claude Opus 5.5; methodology Additional Details · reviewed 2026-09-30	Published configuration Effort: Not specified	lab self-report Provider-published claim. Independent source results are admitted separately with exact checkpoint, effort and compatible protocol; this grid does not establish a matched comparison.	Reviewed 2026-09-30
ProgramBench · 166 golden-task subset	Supporting evidence	91.2% Reported settings & source mini-swe-agent without the six-hour timeout; 34 tasks with a reference binary below 0.9 excluded; scored only on tests the reference binary passes; context up to 1M. Section 8.10.1 does not state reasoning effort. Provider-published lab self-report from the Claude Opus 5.5 system card (SHA256 7311c9c6bbb16d012f1c12c7418b05949fcf7ae3e30d2c40f22050074b2a7378). Not an independent board. Hidden-test pass rate, not the fraction of completely solved programs. Effort is not stated in this section. Comparison limit: Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison. Claude Opus 5.5 System Card · section 8.10.1 · reviewed 2026-09-22	Published configuration Effort: Not specified	lab self-report Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison.	Reviewed 2026-09-22
RiemannBench	Supporting evidence	69.6% Reported settings & source Comparator maximum available thinking/reasoning when reported, otherwise best available result; exact setting not identified by this grid. Results sourced from the official Surge public leaderboard. Provider-published launch claim, not independent measurement by Google for every row. Results sourced from the official Surge public leaderboard. Methodology https://deepmind.google/models/evals-methodology/gemini-4-argon, SHA256 ff1df6bdeddc4c0f48840c09a8a7813b10ae7ad9e053892da68dabf64f09bff7. Preserve source rounding. Methodology says capabilities as of September 2026 and results as of October 2026; review date is source retrieval, not evaluation date. Comparison limit: Provider-published claim. Independent source results are admitted separately with exact checkpoint, effort and compatible protocol; this grid does not establish a matched comparison. Google Gemini 4 Argon launch and evaluation methodology · Launch grid row 12: RiemannBench / Claude Opus 5.5; methodology Additional Details · reviewed 2026-09-30	Published configuration Effort: Not specified	lab self-report Provider-published claim. Independent source results are admitted separately with exact checkpoint, effort and compatible protocol; this grid does not establish a matched comparison.	Reviewed 2026-09-30
SWE-bench Multilingual	Coding	93.9% Reported settings & source Adaptive thinking at max effort; default sampling; five-trial mean; context at most 1M. 300 problems across nine programming languages. Provider-published lab self-report from the Claude Opus 5.5 system card (SHA256 7311c9c6bbb16d012f1c12c7418b05949fcf7ae3e30d2c40f22050074b2a7378). Not an independent board. Launch grid shows 93.9%. Comparison limit: Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison. Claude Opus 5.5 System Card · Table 8.1.A; section 8.2 · reviewed 2026-09-22	Published configuration Effort: Max	lab self-report Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison.	Reviewed 2026-09-22
SWE-bench Multimodal	Supporting evidence	61.4% Reported settings & source Adaptive thinking at max effort; default sampling; five-trial mean; context at most 1M. Visual context added to issue descriptions. Provider-published lab self-report from the Claude Opus 5.5 system card (SHA256 7311c9c6bbb16d012f1c12c7418b05949fcf7ae3e30d2c40f22050074b2a7378). Not an independent board. Comparison limit: Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison. Claude Opus 5.5 System Card · Table 8.1.A; section 8.2 · reviewed 2026-09-22	Published configuration Effort: Max	lab self-report Provider-published lab self-report. Shown as a labeled claim and not admitted as a matched board comparison.	Reviewed 2026-09-22