Across 41 comparable benchmarks, GPT-5.5 has better reported results on 27 and worse on 14. 2 further shared results are listed below, each with the reason it is not counted.
Moderate evidence41 of 43 shared results countedHow we compare
Price basis: Standard rates vs listed rates. The price section below shows the applicable periods.
| Fact | Claude Sonnet 5 | GPT-5.5 |
|---|---|---|
| Context window | 1M tokens | 1M tokens |
| Input / 1M | $2.00 | $5.00 |
| Output / 1M | $10.00 | $30.00 |
| Blended (4:1) | $2.30 | $6.76 |
| Status | Available | Available |
| Released | June 30, 2026 | April 23, 2026 |
Blended is 4:1 input to output with 90% cache hits at listed rates — set your own mix.
Mean of 27 shared benchmarks grouped by capability on a 0–100 scale. How averages are built
GPT-5.5 ahead on 7 of 9 counted
GPT-5.5 ahead on 1 of 1 counted
No fixed-scale average — see the rows in Benchmark results
GPT-5.5 ahead on 2 of 2 counted
GPT-5.5 ahead on 5 of 5 counted
GPT-5.5 ahead on 2 of 3 counted
Claude Sonnet 5 ahead on 1 of 1 counted
GPT-5.5 ahead on 1 of 1 counted
GPT-5.5 ahead on 2 of 2 counted
Claude Sonnet 5 ahead on 7 of 12 counted
No shared comparable results for Cybersecurity.
Published scores, with matching modes where available. Gaps show Claude Sonnet 5 relative to GPT-5.5. Reading the results
| Benchmark | Claude Sonnet 5 | GPT-5.5 | Gap |
|---|---|---|---|
| APEX-SWERepository Engineering · Pass@1 | 46.4Max · tools | 37.0≠Extra-High · tools | +9.4 pts#23 vs #38 of 61 |
| Code MigrationMaintenance & Optimization · Hidden-test pass rate (%) | 44.4Max · tools | 45.2≠Extra-High · tools | −0.8 pts#24 vs #19 of 79 |
| DeepSWERepository Engineering · Pass@1 (DeepSWE v1.1) | 54.0Deep Thinking · tools | 67.0≠Extra-High · tools | −13.0 pts#55 vs #36 of 73 |
| FrontierCode 1.1Code Generation & Editing · Weighted rubric score (%) | 56.2Extra-High · tools· frontiercode-1-1-extended | 56.7Extra-High · tools· frontiercode-1-1-extended | −0.5 pts#11 vs #8 of 51 |
| GSOMaintenance & Optimization · OPT@1 (%) | 37.3Extra-High · tools | 40.2Extra-High · tools | −3.0 pts#9 vs #8 of 24 |
| Terminal-Bench 2.1Agentic Development · Accuracy | 80.4Extra-High · tools | 83.1Extra-High · tools | −2.7 pts#37 vs #33 of 68 |
| Terminal-Bench 4.0Agentic Development · Resolution rate (%) | 12.4Max · tools | 14.6≠Extra-High · tools | −2.2 pts#59 vs #54 of 90 |
| Vibe Code Bench v1.1Code Generation & Editing · Pass rate (%) | 81.3Max · tools | 69.8≠Extra-High · tools | +11.5 pts#31 vs #51 of 121 |
| WeirdML v2ML Engineering · Average accuracy across 17 tasks (%) | 68.8Thinking High · tools | 84.9≠Extra-High · tools | −16.1 pts#14 vs #4 of 39 |
| SimpleBenchCommonsense · Score (AVG@5) | 60.6Normal · no tools | 69.0Normal · no tools | 0.88×#46 vs #31 of 106 |
| Vending-Bench 2Logical Planning · Money Balance (USD)Not counted · Speed and cost telemetry is reported, not scored | 6378Unspecified Reasoning · tools | 7524Unspecified Reasoning · tools | 0.85×#26 vs #22 of 64 |
| FrontierMath Tier 4 v2Mathematics · Accuracy (verification_code) | 29.3Max · tools | 72.5≠Extra-High · tools | −43.2 pts#37 vs #16 of 73 |
| FrontierMath v2Mathematics · Accuracy (verification_code) | 65.6Max · tools | 85.3≠Extra-High · tools | −19.6 pts#38 vs #16 of 85 |
| CritPtScientific Reasoning · Score | 16.9Max · no tools | 27.1≠Extra-High · no tools | 0.62×#36 vs #18 of 134 |
| GPQA DiamondScientific Reasoning · Accuracy | 90.5Extra-High · no tools | 94.0Extra-High · no tools | −3.5 pts#45 vs #19 of 200 |
| SciCodeScientific Computing · Score | 54.3Thinking High · no tools | 56.1Thinking High · no tools | 0.97×#34 vs #28 of 98 |
| MedCodeClinical Workflows · Accuracy (%) | 47.5Max · tools | 49.1≠Extra-High · tools | −1.6 pts#32 vs #26 of 97 |
| MedScribeClinical Workflows · Rubric pass rate (%) | 76.0Max · tools | 86.9≠Extra-High · tools | −10.8 pts#71 vs #20 of 99 |
| APEX-AgentsTool Orchestration · Score | 54.5Reported best | 55.1Reported best | 0.99×#29 vs #27 of 61 |
| OSWorld-VerifiedDesktop Workflows · Accuracy | 81.2Extra-High · tools | 78.7≠Thinking High · tools | +2.5 pts#8 vs #13 of 36 |
| Toolathlon-VerifiedTool Orchestration · Score | 71.6Max · tools | 73.5≠Extra-High · tools | 0.97×#22 vs #16 of 39 |
| BrowseCompFact Finding · Accuracy | 84.7Thinking · tools + web | 84.4≠Thinking High · tools + web | +0.3 pts#13 vs #14 of 61 |
| Context ArenaMemory & Persistence · Accuracy (8 needles, 4K-128K context) | 79.5Max · no tools | 94.2≠Extra-High · no tools | −14.7 pts#23 vs #5 of 61 |
| GDP.pdfDocuments & Charts · Task pass rate (%) | 13.2Max · no tools | 22.6≠Thinking High · no tools | −9.4 pts#48 vs #26 of 96 |
| MMMU-ProVisual Understanding · Accuracy | 77.3Max · no tools | 81.2≠Thinking Medium · no tools | −3.9 pts#49 vs #26 of 136 |
| Creative WritingWriting | 1798Normal · no tools | 1845Normal · no tools | 0.97×#26 vs #17 of 121 |
| AA-AnalystAgentData Analysis · pass^5 (%) | 46.3Max · tools + web | 50.0≠Deep Thinking · tools + web | −3.8 pts#17 vs #11 of 39 |
| AutomationBenchOffice & Business · Pass Rate | 10.7Max · tools· zapier-private-heldout | 16.6≠Extra-High · tools· zapier-private-heldout | −5.9 pts#50 vs #44 of 64 |
| EMB (Excel Modeling)Finance · Accuracy (%) | 66.3Max · tools | 64.5≠Extra-High · tools | +1.8 pts#24 vs #28 of 78 |
| Finance Agent v2Finance · Score | 53.9Max · tools | 51.8≠Extra-High · tools | 1.04×#29 vs #37 of 75 |
| Harvey Lab-AALegal · Criterion pass rate (some entries report all-pass rate) | 90.1Max · tools | 86.3≠Extra-High · tools | +3.8 pts#10 vs #14 of 39 |
| Harvey's Legal Agent BenchmarkLegal · Accuracy (%) | 5.0Max · tools | 3.8≠Extra-High · tools | +1.3 pts#28 vs #31 of 52 |
| Legal Research BenchLegal · Score | 41.8Max · tools | 40.4≠Extra-High · tools | 1.04×#30 vs #36 of 91 |
| Public Benefits Bench v1.1Service Workflows · Criterion pass rate (%) | 66.0Max · tools | 60.9≠Extra-High · tools | +5.1 pts#20 vs #32 of 50 |
| SAGEService Workflows · Accuracy (%) | 48.9Max · tools | 51.5≠Extra-High · tools | −2.6 pts#24 vs #13 of 82 |
| Tax Agent BenchFinance · Accuracy (%, rubric × citation quality) | 62.3Max · tools | 60.5≠Extra-High · tools | +1.8 pts#16 vs #18 of 74 |
| τ³-BankingService Workflows · Score | 37.3Max · tools | 44.6≠Extra-High · tools | 0.84×#30 vs #15 of 120 |
| HLEKnowledge Exams · Accuracy | 57.4Extra-High · tools | 52.2≠Thinking High · tools | +5.2 pts#21 vs #40 of 151 |
| ECICapability Indices · ECI score (capability index, higher is better) | 156Reported best | 159Reported best | 0.98×#22 vs #10 of 168 |
| LiveBenchCross-capability Suites · Accuracy | 76.0Extra-High · no tools | 80.2Extra-High · no tools | −4.1 pts#29 vs #10 of 106 |
| LiveBench 2.0Cross-capability Suites · Overall score (equal-weight mean of 7 suite scores, 0-100) | 73.0Max · task tools | 68.7≠Extra-High · tools | 1.06×#23 vs #28 of 51 |
| Text Arena (Coding)Preference Arenas · Arena Score | 1544Thinking High · no tools | 1505≠Extra-High · no tools | 1.03×#9 vs #17 of 28 |
| Vals IndexCapability IndicesNot counted · This metric has no better direction | 59.6Max · tools | 57.4≠Extra-High · tools | — |
Repository Engineering · Pass@1
Maintenance & Optimization · Hidden-test pass rate (%)
Repository Engineering · Pass@1 (DeepSWE v1.1)
Code Generation & Editing · Weighted rubric score (%)
Maintenance & Optimization · OPT@1 (%)
Agentic Development · Accuracy
Agentic Development · Resolution rate (%)
Code Generation & Editing · Pass rate (%)
ML Engineering · Average accuracy across 17 tasks (%)
Commonsense · Score (AVG@5)
Logical Planning · Money Balance (USD)
Not counted · Speed and cost telemetry is reported, not scored
Mathematics · Accuracy (verification_code)
Mathematics · Accuracy (verification_code)
Scientific Reasoning · Score
Scientific Reasoning · Accuracy
Scientific Computing · Score
Clinical Workflows · Accuracy (%)
Clinical Workflows · Rubric pass rate (%)
Tool Orchestration · Score
Desktop Workflows · Accuracy
Tool Orchestration · Score
Fact Finding · Accuracy
Memory & Persistence · Accuracy (8 needles, 4K-128K context)
Writing
Data Analysis · pass^5 (%)
Office & Business · Pass Rate
Finance · Accuracy (%)
Finance · Score
Legal · Criterion pass rate (some entries report all-pass rate)
Legal · Accuracy (%)
Legal · Score
Service Workflows · Criterion pass rate (%)
Service Workflows · Accuracy (%)
Finance · Accuracy (%, rubric × citation quality)
Service Workflows · Score
Knowledge Exams · Accuracy
Capability Indices · ECI score (capability index, higher is better)
Cross-capability Suites · Accuracy
Cross-capability Suites · Overall score (equal-weight mean of 7 suite scores, 0-100)
Preference Arenas · Arena Score
Capability Indices
Not counted · This metric has no better direction
A dash means no result is available.
| Benchmark | Claude Sonnet 5 | GPT-5.5 |
|---|---|---|
| AIME 2026 | — | 100.0 |
| FrontierMath | — | 51.7 |
| FrontierMath - Tier 4 | — | 35.4 |
| FrontierMath Erdős | — | 0.0 |
| HLE-Verified | 31.0 | — |
| HMMT Feb 2026 | — | 98.5 |
| IMO-ProofBench Advanced | — | 71.9 |
| MathArena Apex | — | 80.2 |
| ProofBench v1.1 (Lean 4) | 77.0 | — |
| Riemann-bench | — | 55.2 |
| CursorBench 4.0 | 34.1 | — |
| FrontierCode 2.0 | — | 22.6 |
| FrontierSWE v1 | — | 64.5 |
| IOI (Vals v2) | 45 | — |
| Kimi Code Bench 2.0 | — | 69.0 |
| PostTrain Bench | — | 27.2 |
| SWE-Atlas-QnA | — | 45.4 |
| SWE-Bench Pro - Public | — | 58.6 |
| SWE-Bench Pro V2 | 93.2 | — |
| SWE-Bench Pro V2 Hard | 88.2 | — |
| SWE-bench Verified | 85.2 | — |
| Terminal Bench 2.0 | — | 82.7 |
| Terminal Bench Hard | — | 60.6 |
| Terminal-Bench 4.1 | 12.1 | — |
| Vibe Code Bench 1-100 | 13.8 | — |
| Agents' Last Exam v1.0 | — | 26.6 |
| MCP-Atlas | — | 75.3 |
| MCPMark-Verified | — | 92.9 |
| OSWorld 2.0 | — | 49.5 |
| PinchBench v2 | — | 75.5 |
| ExploitBench | — | 72.3 |
| Agent Arena | 2 | — |
| Agent Arena Index | 1110 | — |
| Vals Index v2.1 | 51.8 | — |
| AA-LCR | — | 84.3 |
| EBR-bench | — | 34.3 |
| GeoBench ACW | 74.0 | — |
| PerceptionBench | — | 55.8 |
| DECK-Bench | — | 68.2 |
| ARC-AGI-1 | — | 95.0 |
| ARC-AGI-2 | — | 85.0 |
| ARC-AGI-3 | — | 0.43 |
| HealthBench | — | 56.5 |
| HealthBench Professional | — | 51.8 |
Time-of-day providers use peak-hour rates; fixed-price providers use standard rates. Off-peak discounts remain in the published details.
Drag the handle above the bar for cache hits, the one below it for the input : output mix · 5% steps
Blended price = effective input rate at 90% cache hits × 80% + output rate × 20%
Claude Sonnet 5 costs66%lessthan GPT-5.5 per million tokens — 2.9× cheaper.Using standard rates vs listed rates at this token mix.
Claude Sonnet 5: Standard pricing. Anthropic official pricing (platform.claude.com/docs/en/about-claude/pricing) verified 2026-11-13; Promotional introductory pricing extended permanently per Anthropic pricing docs
GPT-5.5: Standard pricing. context_length <= 272000 OpenAI official pricing (developers.openai.com/api/docs/pricing) verified 2026-10-21; standard tier; <=272K input tokens context_length > 272000 OpenAI official pricing (developers.openai.com/api/docs/pricing) verified 2026-10-21; standard tier; >272K input tokens; start date not published, verification date used
| Price / 1M tokens | Claude Sonnet 5 | GPT-5.5 | Ratio |
|---|---|---|---|
| Input | $2.00 | $5.00 | 0.40× |
| Output | $10.00 | $30.00 | 0.33× |
| Cached input | $0.20 | $0.50 | 0.40× |
| Cache write | — | — | — |
| Blended | $2.30 | $6.76 | 0.34× |
Estimates use the selected token mix. Cache writes, tool fees and long-context surcharges are excluded.
Reported cost per task, paired with the score from the same run. Settings and sources may differ; costs do not affect the verdict.
Claude Sonnet 5 relative to GPT-5.5
Summary: 3 of 4 shared benchmarks have matching recorded settings and sources. The average is a geometric mean of 3 positive-cost ratios, one equal weight per benchmark. This benchmark set can include related tests and indexes; it is not a workload cost estimate.
Claude Sonnet 5 1.7× more expensive
Claude Sonnet 5 1.7× cheaper
Claude Sonnet 5 1.5× more expensive
Settings or sources differ
Claude Sonnet 5 1.7× more expensive
Claude Sonnet 5 1.7× cheaper
Claude Sonnet 5 1.5× more expensive
Settings or sources differ
Effort, protocol, aggregation and source match; other run conditions may differ.
Claude Sonnet 5
GPT-5.5
Effort, protocol, aggregation and source match; other run conditions may differ.
Claude Sonnet 5
GPT-5.5
Effort, protocol, aggregation and source match; other run conditions may differ.
Claude Sonnet 5
GPT-5.5
Settings or sources differ or are not recorded; interpret the comparison with care.
Claude Sonnet 5
GPT-5.5
| Benchmark | Claude Sonnet 5 | GPT-5.5 |
|---|---|---|
| AA-LCR | Not reported | |
| Agent Arena | $0.8542 / task Median · Net improvement index (percentage points vs. average model): 2 Thinking High (With Tools + Internet) | Not reported |
| Agent Arena Index | $0.9521 / task Median · Agent Arena Index rating (index points): 1110 Thinking High (With Tools + Internet) | Not reported |
| Agents' Last Exam v1.0 | Not reported | |
| AIME 2026 | Not reported | |
| ARC-AGI-1 | Not reported | 1 more settings |
| ARC-AGI-2 | Not reported | 2 more settings |
| CursorBench 4.0 | $3.48 / task Average · Task score: 30.8 Thinking High (With Tools) · cursorbench-4-0-cursor-agent $1.39 / task Average · Task score: 24.1 Thinking Low (With Tools) · cursorbench-4-0-cursor-agent 2 more settings$2.31 / task Average · Task score: 28.0 Thinking Medium (With Tools) · cursorbench-4-0-cursor-agent $4.55 / task Average · Task score: 32.0 Extra-High (With Tools) · cursorbench-4-0-cursor-agent | Not reported |
| FrontierCode 2.0 | Not reported | $3.8021 / task Average · Weighted rubric score: 19.1 Thinking High (With Tools + Internet) $2.9515 / task Average · Weighted rubric score: 14.2 Thinking Medium (With Tools + Internet) 1 more settings |
| Tax Agent Bench | $1.7948 / task Average · Accuracy (%, rubric × citation quality): 29.4 Unspecified Reasoning (With Tools) | Not reported |
| Terminal-Bench 4.0 | $26.05 / task Average · Resolution rate (%): 10.6 Unspecified Reasoning (With Tools) · vals-mini-swe | Not reported |
| Terminal-Bench 4.1 | Not reported | |
| Vals Index v2.1 | Not reported | |
| Vibe Code Bench 1-100 | Not reported |
| Attribute | Claude Sonnet 5 | GPT-5.5 |
|---|---|---|
| Availability | Available | Available |
| Provider | Anthropic | OpenAI |
| Model type | Multimodal model | Reasoning model |
| First release | June 30, 2026 | April 23, 2026 |
| Context window | 1M tokens | 1M tokens |
| Max output | 128K tokens | 128K tokens |
| Architecture | Undisclosed | Undisclosed |
| Documented capabilities | Reasoning, Multilingual, Coding, Tool use, Function calling, Computer use, Web search, Long context | Reasoning, Multilingual, Coding, Tool use, Function calling, Computer use, Web search, Long context |
| Knowledge cutoff | 2026-01 | — |
| Input modalities | Text, Image | Text, Image |
| Output modalities | Text | Text |
| Weights | Closed / API only | Closed / API only |
| Commercial use | Proprietary | Proprietary |
These are reported-score comparisons, not controlled head-to-head tests.
What people found comparing these models, or using either one. Comments also appear on the models' own pages.
Used Claude Sonnet 5 or GPT-5.5? Sign in with Google to share what worked, what did not, and which API or setup you tested.
No community comments on either model yet.
Alternatives with the most benchmark overlap for this matchup.