AnthropicClaude Haiku 4.5
Claude Haiku 4.5 is an Anthropic proprietary reasoning model, released 15 Oct 2025. Its best configuration (no reasoning reasoning effort) ranks #424 of 785 on the BenchLeader Index at 48.5 ±5.4, in the lower half. The ± is the point: 349 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in knowledge (58) and lowest in multimodal (38). At $2.00 per million tokens blended it is pricier than most ranked models. Output speed of 84 tokens per second puts it slower than most, with a first answer in 0.5 s. It has been measured at 5 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 15 Oct 2025.stale: no new result in six months
- Blended price
- $2.00/M
- $1.00 in · $5.00 out
- Output speed
- 84 tok/s
- measured by Artificial Analysis
- First answer
- 0.48 s
- first token 0.48 s
- Context
- 200k
- Full answer
- 6.46 s
- median, reasoning included
- Cost per run
- –
- Released
- 15 Oct 2025
- Overall index49
- Reasoning41
- Agents & tools58
- Knowledge58
- Instruction following45
- Multimodal38
- Long context49
- Composite47
Versions
Anthropic has shipped 6 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.
| Model | Released | Index | Rank |
|---|---|---|---|
| Claude Haiku 5.5xhigh | 7 Oct 2026 | 61.5 | #95 |
| Claude Haiku 4.5no reasoningthis page | 15 Oct 2025 | 48.5 | #424 |
| Claude 3.5 Haiku | 4 Nov 2024 | 39.7 | #675 |
| Claude 3 Haiku | 7 Mar 2024 | 37.0 | #738 |
| Claude Haiku 3 | – | – | – |
| Claude Haiku 4thinking | – | – | – |
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| no reasoningbest | 48.5 | #424 | 84 tok/s | 0.48 s | $0.0019 | Agents & tools 58 · Composite 47 · Instruction following 45 · Knowledge 58 · Long context 49 · Multimodal 38 · Reasoning 41 |
| low | – | – | 63 tok/s | 0.87 s | $0.0019 | |
| high | – | – | 63 tok/s | 0.87 s | $0.0019 | Coding 59 |
| thinking | 46.7 | #476 | 94 tok/s | 19 s | $0.0019 | Agents & tools 41 · Coding 27 · Composite 48 · Instruction following 47 · Knowledge 52 · Long context 62 · Maths 58 · Multimodal 41 · Reasoning 39 |
| not stated | 45.3 | #516 | 63 tok/s | 0.87 s | $0.0019 | Agents & tools 45 · Coding 48 · Composite 26 · Human preference 49 · Instruction following 52 · Knowledge 35 · Maths 61 · Reasoning 43 |
How its index has moved
8 Sept 2026 to 25 Oct 2026- no reasoning
- thinking
- not stated
The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| GPQA Diamondeffort not stated | – | – | 71.2%#168 | Epoch AI Benchmarking Hub | |
| Arena Hard Promptseffort not stated | – | – | 1442#121 | Arena (LMArena) | |
| LiveBench 2.0 Reasoningnot in indexeffort not stated | – | – | 61.7%#51 | LiveBench | |
| GPQA Diamond (AA)not in index | 64.7%#345 | 67.2%#322 | – | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 4.2%#522 | 10.4%#318 | – | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | 72.2%#84 | – | Vals AI | |
| GPQA Diamond (HELM)not in indexeffort not stated | – | – | 60.5%#27 | HELM Capabilities | |
| ARC-AGI-1 | – | 47.7%#183 | 14.3%#229 | ARC Prize | |
| ARC-AGI-2 | – | 4.0%#185 | 1.3%#218 | ARC Prize | |
| CritPt | 0.0%#315 | 0.0%#315 | – | Artificial Analysis | |
| Chess Puzzlesnot in indexeffort not stated | – | – | 8.0%#124 | Epoch AI Benchmarking Hub | |
| LMCAnot in indexeffort not stated | – | – | 30.9%#134 | Epoch AI Benchmarking Hub | |
| DTBenchnot in indexeffort not stated | – | – | 73.6%#153 | Epoch AI Benchmarking Hub | |
| ForecastBenchnot in indexeffort not stated | – | – | 63.4%#40 | ForecastBench |
Coding
Agents & tools
Maths
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| OTIS Mock AIMEeffort not stated | – | 66.7%#149 | Epoch AI Benchmarking Hub | |
| MATH Level 5effort not stated | – | 96.4%#11 | Epoch AI Benchmarking Hub | |
| AIME (Vals) | 82.7%#44 | – | Vals AI | |
| MGSM | 92.2%#21 | – | Vals AI | |
| Omni-MATH (HELM)effort not stated | – | 56.1%#12 | HELM Capabilities | |
| Arena Mathseffort not stated | – | 1402#147 | Arena (LMArena) |
Knowledge
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| SimpleQA Verifiedeffort not stated | – | – | 13.2%#80 | Epoch AI Benchmarking Hub | |
| LiveBench 2.0 Data Analysisnot in indexeffort not stated | – | – | 59.3%#47 | LiveBench | |
| LiveBench Researchnot in indexeffort not stated | – | – | 42.2%#50 | LiveBench | |
| AA-Omniscience | -7.6#176 | -4.4#159 | – | Artificial Analysis | |
| MMLU-Pro | – | 78.7%#95 | – | Vals AI | |
| LegalBench | – | 81.2%#81 | – | Vals AI | |
| CorpFin | – | 60.6%#66 | 60.3%#69 | Vals AI | |
| TaxEval | – | 67.5%#102 | – | Vals AI | |
| MedQA | – | 79.6%#70 | – | Vals AI | |
| MMLU-Pro (HELM)not in indexeffort not stated | – | – | 77.7%#27 | HELM Capabilities | |
| CaseLaw v2not in index | – | 56.5%#30 | – | Vals AI | |
| Excel Modeling Benchmarknot in index | – | 22.7%#67 | – | Vals AI | |
| MedCodenot in index | – | 32.7%#85 | – | Vals AI | |
| MedScribenot in index | – | 85.2%#30 | – | Vals AI | |
| Public Benefits Bench v1.1not in indexeffort not stated | – | – | 54.3%#40 | Vals AI | |
| AA-Omniscience: accuracynot in index | 14.4%#436 | 18.0%#346 | – | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 74.3%#26 | 72.7%#33 | – | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| IFBench | 42.0%#239 | 54.3%#154 | – | Artificial Analysis | |
| MultiChallenge | – | 50.5%#27 | – | Scale AI SEAL | |
| IFEval (HELM)effort not stated | – | – | 80.1%#48 | HELM Capabilities | |
| WildBench (HELM)effort not stated | – | – | 83.9%#15 | HELM Capabilities | |
| Arena Instruction Followingnot in indexeffort not stated | – | – | 1415#114 | Arena (LMArena) |
Human preference
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Text | 1415#133 | Arena (LMArena) | |
| EQ-Bench 4 | 1057#28 | EQ-Bench | |
| Arena Creative Writingnot in index | 1387#126 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1425#118 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1437#110 | Arena (LMArena) |
Multimodal
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| LiveBench Designnot in indexeffort not stated | – | – | 59.9%#48 | LiveBench | |
| MMMU-Pro | 55.1%#228 | 58.5%#217 | – | Artificial Analysis | |
| MMMU-Pro (Vals)not in index | – | 46.1%#88 | – | Vals AI | |
| MortgageTaxnot in index | – | 62.2%#54 | – | Vals AI | |
| SAGEnot in index | – | 31.8%#74 | – | Vals AI | |
| Blueprint-Bench 2not in indexeffort not stated | – | – | 0.0%#29 | Epoch AI Benchmarking Hub |
Long context
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| AA-LCR | 49.7%#325 | 74.3%#162 | – | Artificial Analysis | |
| Arena Documentnot in indexeffort not stated | – | – | 1419#42 | Arena (LMArena) | |
| MLCRnot in index | – | 5.0%#145 | – | Artificial Analysis |
Composite
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | – | 142.2#103 | Epoch AI Benchmarking Hub | |
| LiveBench 2.0effort not stated | – | – | 46.0%#49 | LiveBench | |
| AA Intelligence Index v4.3.2 | 15.4#283 | 16.9#267 | – | Artificial Analysis | |
| HELM Capabilities meannot in indexeffort not stated | – | – | 71.7%#22 | HELM Capabilities | |
| Vals Multimodal Indexnot in index | – | 42.9%#28 | – | Vals AI |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| Google Vertex | 153 tok/s | 2.58 s | $1.10 | $5.50 | 200k | – |
| Amazon Bedrock | 130 tok/s | 0.65 s | $1.10 | $5.50 | 200k | – |
| Amazon Bedrock | 112 tok/s | 0.56 s | $1.10 | $5.50 | 200k | – |
| Google Vertex | 104 tok/s | 0.57 s | $1.10 | $5.50 | 200k | – |
| Google Vertex | 91 tok/s | 0.69 s | $1.00 | $5.00 | 200k | – |
| Azure | 79 tok/s | 0.44 s | $1.00 | $5.00 | 200k | – |
| Anthropic | 67 tok/s | 0.44 s | $1.00 | $5.00 | 200k | – |
| Amazon Bedrock | 60 tok/s | 0.81 s | $1.00 | $5.00 | 200k | – |
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.100 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0019 | $0.0016 | 4.1 s |
| Summarize a 30-page report | 12,000 / 600 | $0.015 | $0.0069 | 7.7 s |
| Code edit | 6,000 / 1,500 | $0.013 | $0.0095 | 18.4 s |
| Agentic coding session | 60,000 / 4,000 | $0.080 | $0.040 | 48.3 s |
| Structured extraction | 2,000 / 200 | $0.0030 | $0.0017 | 2.9 s |
More from Anthropic
See also
Questions people ask
- What are Claude Haiku 4.5's benchmark scores?
- As of 25 Oct 2026, Claude Haiku 4.5 (no reasoning effort) scores 48.5 ±5.4 on the BenchLeader Index, strongest in knowledge (58) and weakest in multimodal (38). That rests on 7 independent benchmark results, and every score on this page links to its publisher.
- How does Claude Haiku 4.5 rank?
- Claude Haiku 4.5 (no reasoning effort) is #424 of 785 ranked configurations on 25 Oct 2026. 349 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does Claude Haiku 4.5 cost per 1M tokens?
- As of 25 Oct 2026, Claude Haiku 4.5 costs $1.00 per 1M input tokens and $5.00 per 1M output tokens, $2.00 blended (three input tokens per output token).
- When was Claude Haiku 4.5 released?
- Anthropic released Claude Haiku 4.5 on 15 Oct 2025. BenchLeader has results for it at 5 reasoning-effort settings.
Data as of 25 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Claude Haiku 4.5: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/claude-haiku-4-5.