Updated daily at 05:00 UTC
Vision-language
Claude Haiku 4.5
- Released
- 15 Oct 2025
- Status
- Current
- Context
- 200K
- Max output
- 64K
- Input per 1M tokens
- $1.00
- Output per 1M tokens
- $5.00
- Open weights
- No
- License
- proprietary
- Parameters
- —
- Knowledge cutoff
- Feb 2025
- Reasoning
- Yes
- Modalities
- text, image → text
Category scores
Out of 100, from independent and third-party results only, by method category-2026-09-29.1.
| Category | Score | Rank | Coverage |
|---|---|---|---|
| Knowledge & reasoning | 19.8 | 79 | 2 of 4 |
| Agentic tool use | 38.4 | 8 | 2 of 2 |
| Factuality & hallucination | 13.2 | 63 | 1 of 1 |
| Human preference | 48.2 | 32 | 3 of 5 |
Benchmark results
Every result we hold for this model, with who ran it and where it was published.
Knowledge & reasoning
| Benchmark | Metric | Value | Provenance | Setting | Evaluated | Source |
|---|---|---|---|---|---|---|
| Chess Puzzles | Puzzles solved | 8.0% | Independent | 32K | 16 Jul 2026 | Epoch AI |
| DTBench (Decision Theory Benchmark) | Accuracy (capability questions) | 73.6% | Third party | unknown | — | https://conceptualreasoning.ai/dtbench via Epoch AI |
| LMCA (Language Model Conceptual Argumentation) | Correlation with expert ratings (Pearson r) | 0.3094 | Third party | unknown | — | https://conceptualreasoning.ai/lmca via Epoch AI |
Scientific reasoning
| Benchmark | Metric | Value | Provenance | Setting | Evaluated | Source |
|---|---|---|---|---|---|---|
| GPQA Diamond | Accuracy | 71.2% | Independent | 32K | 22 Oct 2025 | Epoch AI |
| GPQA Diamond | Accuracy | 60.5% | Independent | — | 16 Oct 2025 | Epoch AI |
Mathematics
| Benchmark | Metric | Value | Provenance | Setting | Evaluated | Source |
|---|---|---|---|---|---|---|
| FrontierMath Tier 4 (v1) | Accuracy | 2.1% | Independent | 32K | 22 Oct 2025 | Epoch AI |
| FrontierMath Tiers 1-3 (v1) | Accuracy | 5.9% | Independent | 32K | 22 Oct 2025 | Epoch AI |
| FrontierMath Tiers 1-3 (v1) | Accuracy | 4.1% | Independent | — | 16 Oct 2025 | Epoch AI |
| MATH Level 5 | Accuracy | 96.4% | Independent | 32K | 22 Oct 2025 | Epoch AI |
| MATH Level 5 | Accuracy | 86.9% | Independent | — | 16 Oct 2025 | Epoch AI |
| OTIS Mock AIME 2024-2025 | Accuracy | 66.7% | Independent | 32K | 22 Oct 2025 | Epoch AI |
| OTIS Mock AIME 2024-2025 | Accuracy | 35.8% | Independent | — | 16 Oct 2025 | Epoch AI |
Coding & software engineering
| Benchmark | Metric | Value | Provenance | Setting | Evaluated | Source |
|---|---|---|---|---|---|---|
| ExploitBench | Mean capability | 13.7% | Source not stated | unknown | — | Epoch AI via Epoch AI |
| ExploitBench | Mean capability | 13.4% | Source not stated | unknown | — | Epoch AI via Epoch AI |
| Terminal-Bench 2.0 | Accuracy (mean over trials) | 35.5% | Third party | unknown | — | https://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI |
| Terminal-Bench 2.0 | Accuracy (mean over trials) | 29.8% | Third party | unknown | — | https://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI |
| Terminal-Bench 2.0 | Accuracy (mean over trials) | 29.8% | Third party | — | — | Terminal-Bench v2 Leaderboard via Epoch AI |
| Terminal-Bench 2.0 | Accuracy (mean over trials) | 28.3% | Third party | unknown | — | https://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI |
| Terminal-Bench 2.0 | Accuracy (mean over trials) | 28.3% | Third party | — | — | Terminal-Bench v2 Leaderboard via Epoch AI |
| Terminal-Bench 2.0 | Accuracy (mean over trials) | 27.5% | Third party | unknown | — | https://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI |
| Terminal-Bench 2.0 | Accuracy (mean over trials) | 27.5% | Third party | — | — | Terminal-Bench v2 Leaderboard via Epoch AI |
| Terminal-Bench 2.0 | Accuracy (mean over trials) | 13.9% | Third party | unknown | — | https://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI |
| Terminal-Bench 2.0 | Accuracy (mean over trials) | 13.9% | Third party | — | — | Terminal-Bench v2 Leaderboard via Epoch AI |
| WeirdML (v2) | Average accuracy | 45.4% | Third party | — | — | https://htihle.github.io/weirdml.html via Epoch AI |
| WeirdML (v2) | Average accuracy | 44.1% | Third party | 16K | — | https://htihle.github.io/weirdml.html via Epoch AI |
Agentic tool use
| Benchmark | Metric | Value | Provenance | Setting | Evaluated | Source |
|---|---|---|---|---|---|---|
| BALROG | Average progress | 31.2% | Third party | 1K | — | https://balrogai.com/ via Epoch AI |
| BALROG | Average progress | 31.2% | Third party | — | — | Balrog Leaderboard via Epoch AI |
| Deep Research Bench (FutureSearch) | Average score | 45.5% | Third party | low | — | https://evals.futuresearch.ai/#drb via Epoch AI |
Factuality & hallucination
| Benchmark | Metric | Value | Provenance | Setting | Evaluated | Source |
|---|---|---|---|---|---|---|
| SimpleQA Verified | Accuracy (% correct) | 13.2% | Independent | — | 10 Aug 2026 | Epoch AI |
| SimpleQA Verified | Accuracy (% correct) | 12.6% | Independent | 32K | 27 Aug 2026 | Epoch AI |
Human preference
| Benchmark | Metric | Value | Provenance | Setting | Evaluated | Source |
|---|---|---|---|---|---|---|
| Code Arena: WebDev (arena.ai) | Arena score (Elo scale) | 1330 | Independent | — | 25 Sep 2026 | Arena |
| Code Arena: WebDev (arena.ai) | Votes | 29211 | Independent | — | 25 Sep 2026 | Arena |
| Document Arena (arena.ai) | Arena score (Elo scale) | 1419 | Independent | — | 13 Sep 2026 | Arena |
| Document Arena (arena.ai) | Votes | 34677 | Independent | — | 13 Sep 2026 | Arena |
| Text Arena (arena.ai) | Arena score (Elo scale) | 1414 | Independent | — | 25 Sep 2026 | Arena |
| Text Arena (arena.ai) | Votes | 140791 | Independent | — | 25 Sep 2026 | Arena |
Other Claude Haiku versions
- Claude 3.5 Haiku4 Nov 2024
- Claude 3 Haiku13 Mar 2024