Model comparison
Claude Haiku 4.5 vs Phi-4
Claude Haiku 4.5 is the stronger model overall, scoring 39.3 to 30.9 on the Noometry Index. Phi-4 costs 23× less per token, which makes it the better buy when Claude Haiku 4.5's lead doesn't matter for your workload.
Last verified . 25 shared benchmarks.
MicrosoftSummary
- They share 25 benchmarks with published results for both. Claude Haiku 4.5 scores higher in 8 categories and Phi-4 in 1 category; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where Claude Haiku 4.5 leads 43.0 to 19.6.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 66.7% for Claude Haiku 4.5 and 13.8% for Phi-4.
- Phi-4 is cheaper at $0.07 / $0.14 per million input/output tokens, against $1 / $5 for Claude Haiku 4.5.
- Claude Haiku 4.5 accepts more context: 200K tokens versus 128K.
- Phi-4 has downloadable open weights; the other is API-only.
Side by side
| Claude Haiku 4.5 | Phi-4 | |
|---|---|---|
| Provider | Anthropic | Microsoft |
| Noometry Index | 39.3 | 30.9 |
| Released | 2025-10-15 | 2024-12-11 |
| Weights | Proprietary | Open |
| Context window | 200K | 128K |
| Max output | 64K | 4K |
| Input $ / M tokens | $1 | $0.07 |
| Output $ / M tokens | $5 | $0.14 |
| Results tracked | 53 | 37 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Claude Haiku 4.5 leads
Claude Haiku 4.5: 43.9 (#71), Phi-4: 33.8 (#237)
| Benchmark | Claude Haiku 4.5 | Phi-4 |
|---|---|---|
| LMArena Coding | 1454 | 1231 |
| SWE-bench Verified (bash only) | 66.6% | — |
| LMArena WebDev | 1330 | — |
| SWE-bench Multilingual | 64.7% | — |
| SciCode | 43.3% | — |
| WeirdML | 45.4% | — |
| BigCodeBench Instruct | — | 45.5% |
| LiveBench Coding | — | 30.7% |
| BigCodeBench Complete | — | 55.4% |
| ALE-Bench | 653.48 | — |
Agentic & Tool Use Claude Haiku 4.5 leads
Claude Haiku 4.5: 33.5 (#50), Phi-4: 22.6 (#120)
| Benchmark | Claude Haiku 4.5 | Phi-4 |
|---|---|---|
| Berkeley Function Calling Leaderboard | 68.7% | 28.8% |
| BALROG | 31.2% | 11.6% |
| Terminal-Bench | 35.5% | — |
| DeepResearch Bench | 45.5% | — |
| ExploitBench | 13.7% | — |
| Vending-Bench 2 | 458.89 | — |
Reasoning Phi-4 leads
Claude Haiku 4.5: 15.1 (#310), Phi-4: 17.0 (#287)
| Benchmark | Claude Haiku 4.5 | Phi-4 |
|---|---|---|
| Chess Puzzles | 8% | 1% |
| LMArena Hard Prompts | 1420 | 1220 |
| Epoch Capabilities Index | 142.42 | 130.43 |
| ARC-AGI-2 | 4% | — |
| NYT Connections (extended) | 14.3% | — |
| ARC-AGI-1 | 47.7% | — |
| CritPt | 0% | — |
| LiveBench Reasoning | — | 47.8% |
| DTBench | 73.6% | — |
| LiveBench Data Analysis | — | 45.2% |
| LMCA | 30.9% | — |
| ForecastBench | 61.4 | — |
| LiveBench | — | 41.6% |
Math Claude Haiku 4.5 leads
Claude Haiku 4.5: 43.0 (#69), Phi-4: 19.6 (#279)
| Benchmark | Claude Haiku 4.5 | Phi-4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.7% | 13.8% |
| LMArena Math | 1395 | 1246 |
| MATH Level 5 | 96.4% | 64.9% |
| Omni-MATH | 56.1% | — |
| LiveBench Math | — | 42% |
| FrontierMath (Feb 2025 set) | 5.9% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge Claude Haiku 4.5 leads
Claude Haiku 4.5: 37.6 (#146), Phi-4: 32.5 (#204)
| Benchmark | Claude Haiku 4.5 | Phi-4 |
|---|---|---|
| GPQA Diamond | 71.2% | 56.1% |
| Vectara Hallucination Rate | 9.8% | 3.7% |
| LMArena Expert | 1443 | 1203 |
| SimpleQA Verified | 13.2% | — |
| MMLU-Pro | 77.7% | — |
| Confabulations | — | 29.4% |
| GPQA (HELM) | 60.5% | — |
| MMLU | — | 84.8% |
Multimodal Not comparable
Claude Haiku 4.5: 27.0 (#111), Phi-4: —
| Benchmark | Claude Haiku 4.5 | Phi-4 |
|---|---|---|
| Blueprint-Bench 2 | 0% | — |
| LMArena Document | 1420 | — |
Multilingual Claude Haiku 4.5 leads
Claude Haiku 4.5: 50.1 (#124), Phi-4: 37.4 (#232)
| Benchmark | Claude Haiku 4.5 | Phi-4 |
|---|---|---|
| LMArena Non-English | 1376 | 1197 |
| LMArena Chinese | 1415 | 1211 |
| LMArena French | 1408 | 1223 |
| LMArena German | 1377 | 1222 |
| LMArena Japanese | 1339 | 1158 |
| LMArena Korean | 1346 | 1151 |
| LMArena Russian | 1380 | 1209 |
| LMArena Spanish | 1420 | 1234 |
Instruction Following Claude Haiku 4.5 leads
Claude Haiku 4.5: 71.6 (#144), Phi-4: 60.6 (#246)
| Benchmark | Claude Haiku 4.5 | Phi-4 |
|---|---|---|
| LMArena Instruction Following | 1414 | 1201 |
| LiveBench Instruction Following | — | 58.4% |
| IFEval | 80.1% | — |
Long Context Claude Haiku 4.5 leads
Claude Haiku 4.5: 43.6 (#86), Phi-4: 36.9 (#221)
| Benchmark | Claude Haiku 4.5 | Phi-4 |
|---|---|---|
| LMArena Longer Query | 1427 | 1217 |
Writing & Preference Claude Haiku 4.5 leads
Claude Haiku 4.5: 58.1 (#119), Phi-4: 40.7 (#239)
| Benchmark | Claude Haiku 4.5 | Phi-4 |
|---|---|---|
| LMArena Text | 1396 | 1217 |
| LMArena Creative Writing | 1371 | 1182 |
| LMArena Multi-Turn | 1409 | 1206 |
| Short-Story Creative Writing | — | 62.6% |
| WildBench | 83.9% | — |
| EQ-Bench 4 | 1064 | — |
| LiveBench Language | — | 25.6% |
Frequently asked questions
Is Claude Haiku 4.5 better than Phi-4?
Claude Haiku 4.5 is the stronger model overall, scoring 39.3 to 30.9 on the Noometry Index. Phi-4 costs 23× less per token, which makes it the better buy when Claude Haiku 4.5's lead doesn't matter for your workload.
Which is cheaper, Claude Haiku 4.5 or Phi-4?
Phi-4 is cheaper. It lists at $0.07 per million input tokens and $0.14 per million output tokens; Claude Haiku 4.5 lists at $1 and $5.
Is Claude Haiku 4.5 or Phi-4 better for coding?
Claude Haiku 4.5 scores higher on coding benchmarks: 43.9 versus 33.8 in the Noometry coding category.
Which has the bigger context window?
Claude Haiku 4.5 does, with 200K tokens against 128K.
How many benchmarks do Claude Haiku 4.5 and Phi-4 share?
25 benchmarks have published results for both models. Claude Haiku 4.5 has 53 scored results on Noometry and Phi-4 has 37.