Across 25 comparable benchmarks, Claude Sonnet 4 and Haiku 4.5 are evenly matched at 12 apiece. 1 further shared result is listed below, each with the reason it is not counted.
Moderate evidence25 of 26 shared results countedHow we compare
Claude Sonnet 4 is no longer available through its official API. API prices for this version are not shown.
Price basis: Listed rates vs standard rates. The price section below shows the applicable periods.
| Fact | Claude Sonnet 4 | Haiku 4.5 |
|---|---|---|
| Context window | 200K tokens | 200K tokens |
| Input / 1M | — | $1.00 |
| Output / 1M | — | $5.00 |
| Blended (4:1) | — | $1.15 |
| Status | Retired | Available |
| Released | May 23, 2025 | October 15, 2025 |
Blended is 4:1 input to output with 90% cache hits at listed rates — set your own mix.
Mean of 19 shared benchmarks grouped by capability on a 0–100 scale. How averages are built
Claude Sonnet 4 ahead on 3 of 5 counted
Haiku 4.5 ahead on 1 of 1 counted
Haiku 4.5 ahead on 2 of 3 counted
Claude Sonnet 4 ahead on 2 of 3 counted
Haiku 4.5 ahead on 3 of 3 counted
Claude Sonnet 4 ahead on 2 of 2 counted
Claude Sonnet 4 ahead on 4 of 4 counted
No shared comparable results for Research & search, Long context, Cybersecurity.
Published scores, with matching modes where available. Gaps show Claude Sonnet 4 relative to Haiku 4.5. Reading the results
| Benchmark | Claude Sonnet 4 | Haiku 4.5 | Gap |
|---|---|---|---|
| LiveCodeBenchAlgorithmic Coding · Pass @K | 48.5Normal · no tools | 62.0≠Extended · no tools | 0.78×#88 vs #69 of 113 |
| SWE-Bench Pro - PublicRepository Engineering · Accuracy | 42.7Normal · tools | 39.5≠Extended · tools | +3.3 pts#58 vs #63 of 66 |
| SWE-bench VerifiedRepository Engineering · Accuracy | 80.2Parallel Thinking · tools | 73.3≠Thinking · tools · 128K | +6.9 pts#15 vs #47 of 108 |
| Terminal Bench HardAgentic Development · Accuracy | 31.1Thinking · tools | 27.3≠Normal · tools | +3.8 pts#66 vs #74 of 157 |
| Terminal-Bench 1.0Agentic Development · Accuracy | 41.3Deep Thinking · tools | 41.8≠Thinking · tools · 32K | −0.5 pts#9 vs #8 of 26 |
| ARC-AGI-1Abstract Generalization · Score (%) | 40.0Thinking · no tools | 47.7≠Extended · no tools | −7.7 pts#69 vs #67 of 87 |
| ARC-AGI-2Abstract Generalization · Score (% solved); cost per task (USD)Not counted · Speed and cost telemetry is reported, not scored | 5.9Thinking · no tools | 4.5≠Extended · no tools | 1.32×#64 vs #70 of 80 |
| AIME2025Mathematics · Accuracy | 85.0Deep Thinking · tools | 96.3≠Thinking · tools · 128K | −11.3 pts#48 vs #18 of 86 |
| FrontierMathMathematics · Accuracy | 4.1Normal · no tools | 4.1Normal · no tools | =#32 vs #32 of 48 |
| FrontierMath - Tier 4Mathematics · Accuracy | 0.0Normal · no tools | 2.1≠Thinking · no tools · 32K | −2.1 pts#44 vs #31 of 50 |
| GPQA DiamondScientific Reasoning · Accuracy | 83.8Deep Thinking · tools | 73.3≠Extended · no tools | +10.5 pts#91 vs #126 of 200 |
| MedCodeClinical Workflows · Accuracy (%) | 35.0Thinking · tools | 32.7Thinking · tools | +2.3 pts#73 vs #81 of 97 |
| MedScribeClinical Workflows · Rubric pass rate (%) | 72.4Normal · tools | 85.2≠Thinking · tools | −12.8 pts#81 vs #30 of 99 |
| Claw BenchTool Orchestration · Accuracy | 77.8Thinking · tools | 89.4Thinking · tools | −11.6 pts#23 vs #11 of 29 |
| Pinch BenchTool Orchestration · Accuracy | 80.5Thinking · tools | 82.0Thinking · tools | −1.5 pts#23 vs #22 of 38 |
| PinchBench v2Tool Orchestration · Average score (%) | 48.8Reported best | 67.7Reported best | −18.9 pts#45 vs #33 of 51 |
| MMMUVisual Understanding · Accuracy | 74.4Normal · no tools | 73.2≠Thinking · no tools · 128K | +1.2 pts#23 vs #27 of 62 |
| MMMU-ProVisual Understanding · Accuracy | 62.4Normal · no tools | 58.6≠Thinking · no tools | +3.8 pts#100 vs #109 of 136 |
| IF BenchInstruction Following · Accuracy | 55.0Thinking · tools | 54.3≠Extended · no tools | +0.7 pts#34 vs #36 of 38 |
| SAGEService Workflows · Accuracy (%) | 35.0Normal · tools | 31.8≠Thinking · tools | +3.2 pts#67 vs #71 of 82 |
| τ²-BenchService Workflows · Accuracy | 52.0Normal · tools | 33.0Normal · tools | +19.0 pts#32 vs #39 of 40 |
| τ³-BankingService Workflows · Score | 16.7Thinking · tools | 9.3Thinking · tools | 1.80×#67 vs #95 of 120 |
| HLEKnowledge Exams · Accuracy | 9.6Thinking · no tools | 9.7≠Extended · no tools | −0.1 pts#130 vs #129 of 151 |
| MMLU-ProKnowledge Exams · Accuracy | 84.0Thinking · no tools | 80.0≠Normal · no tools | +4.0 pts#42 vs #62 of 153 |
| ECICapability Indices · ECI score (capability index, higher is better) | 142Reported best | 142Reported best | 0.99×#91 vs #90 of 168 |
| LiveBenchCross-capability Suites · Accuracy | 61.3Thinking · no tools · 64K | 61.3Thinking · no tools · 64K | −0.0 pts#83 vs #82 of 106 |
Algorithmic Coding · Pass @K
Repository Engineering · Accuracy
Repository Engineering · Accuracy
Agentic Development · Accuracy
Agentic Development · Accuracy
Mathematics · Accuracy
Mathematics · Accuracy
Mathematics · Accuracy
Scientific Reasoning · Accuracy
Clinical Workflows · Accuracy (%)
Clinical Workflows · Rubric pass rate (%)
Tool Orchestration · Accuracy
Tool Orchestration · Accuracy
Tool Orchestration · Average score (%)
Instruction Following · Accuracy
Service Workflows · Accuracy (%)
Service Workflows · Accuracy
Service Workflows · Score
A dash means no result is available.
| Benchmark | Claude Sonnet 4 | Haiku 4.5 |
|---|---|---|
| AIME 2024 | 43.4 | — |
| CritPt | 1.1 | — |
| IMO-ProofBench | 27.1 | — |
| IMO-ProofBench Advanced | 4.8 | — |
| MATH-500 | 99.1 | — |
| SciCode | — | 42.2 |
| Aider-Polyglot | 61.3 | — |
| Code Migration | — | 7.5 |
| CodeClash | 1223 | — |
| IOI 2024 (Vals v1) | 9.7 | — |
| IOI 2025 (Vals v1) | 4 | — |
| SWE-Bench Pro - Commercial | 9.1 | — |
| SWE-Bench Pro V2 Hard | — | 25.5 |
| Vibe Code Bench v1.1 | — | 11.4 |
| MCP-Atlas | — | 40.2 |
| OSWorld 2.1 | — | 15.7 |
| OSWorld-Verified | 42.2 | — |
| ExploitBench | — | 13.7 |
| AA Intelligence Index v5.0.1 | — | 7.4 |
| LiveBench 2.0 | — | 46.0 |
| Vals Index | — | 22.9 |
| Creative Writing | 1483 | — |
| Context Arena | — | 36.3 |
| Fiction.liveBench | 46.9 | — |
| GDP.pdf | — | 3.8 |
| VPCT | 34.0 | — |
| AA-AnalystAgent | — | 15.0 |
| AutomationBench | — | 0.46 |
| EMB (Excel Modeling) | — | 22.7 |
| Finance Agent v2 | — | 31.0 |
| GDPval-AA | 33.0 | — |
| Harvey Lab-AA | — | 61.0 |
| Harvey's Legal Agent Benchmark | — | 0.83 |
| Legal Research Bench | — | 10.6 |
| Public Benefits Bench v1.1 | — | 54.3 |
| BALROG | — | 31.2 |
| SimpleBench | 45.5 | — |
| Vending-Bench 2 | — | 459 |
| GAIA | — | 56.4 |
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%
A blended cost needs a published input and output price on both sides, and one of these two does not have them yet.
Claude Sonnet 4: Listed text rates; pricing tier not confirmed.
Haiku 4.5: Standard pricing. Anthropic official pricing (platform.claude.com/docs/en/about-claude/pricing) verified 2026-11-13
| Price / 1M tokens | Claude Sonnet 4 | Haiku 4.5 | Ratio |
|---|---|---|---|
| Input | — | $1.00 | — |
| Output | — | $5.00 | — |
| Cached input | — | $0.10 | — |
| Cache write | — | — | — |
| Blended | — | $1.15 | — |
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 4 relative to Haiku 4.5
No matching recorded settings for a cost summary.
Summary: 0 of 1 shared benchmarks have matching recorded settings and sources. The average is a geometric mean of 0 positive-cost ratios, one equal weight per benchmark. This benchmark set can include related tests and indexes; it is not a workload cost estimate.
Settings or sources differ or are not recorded; interpret the comparison with care.
Claude Sonnet 4
Haiku 4.5
| Benchmark | Claude Sonnet 4 | Haiku 4.5 |
|---|---|---|
| AA Intelligence Index v5.0.1 | Not reported | |
| ARC-AGI-2 | Not reported | |
| AutomationBench | Not reported | $0.07 / task Average · Pass Rate: 0.46 Reported best (effort unspecified) · zapier-private-heldout |
| LiveBench 2.0 | Not reported | $0.934 / task Average · Overall score (equal-weight mean of 7 suite scores, 0-100): 46.0 Max (task-specific tools) |
| Public Benefits Bench v1.1 | Not reported | $0.2245 / task Average · Criterion pass rate (%): 54.3 Unspecified Reasoning (With Tools) |
| Attribute | Claude Sonnet 4 | Haiku 4.5 |
|---|---|---|
| Availability | Retired | Available |
| Provider | Anthropic | Anthropic |
| Model type | Reasoning model | Multimodal model |
| First release | May 23, 2025 | October 15, 2025 |
| Context window | 200K tokens | 200K tokens |
| Max output | 64K tokens | 64K tokens |
| Architecture | Undisclosed | Undisclosed |
| Documented capabilities | Reasoning, Multilingual | Reasoning, Multilingual |
| 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 4 or Haiku 4.5? 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.