MicrosoftPhi-4
Phi-4 is a Microsoft open-weights model, released 12 Dec 2024. Its best configuration ranks #699 of 785 on the BenchLeader Index at 38.9 ±3.8, in the lower half. The ± is the point: 224 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 maths (42) and lowest in long context (23). At $0.223 per million tokens blended it is cheaper than most ranked models. Output speed of 38 tokens per second puts it in the slowest quarter, with a first answer in 2.7 s. Last measured 22 Oct 2026.
- Blended price
- $0.223/M
- $0.130 in · $0.500 out
- Output speed
- 38 tok/s
- measured by Artificial Analysis
- First answer
- 2.67 s
- first token 0.47 s
- Context
- 16k
- Full answer
- 16 s
- median, reasoning included
- Cost per run
- –
- Released
- 12 Dec 2024
- Overall index39
- Reasoning42
- Coding41
- Agents & tools40
- Maths42
- Knowledge36
- Instruction following28
- Human preference40
- Long context23
- Composite36
Versions
Microsoft has shipped 4 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.
How its index has moved
8 Sept 2026 to 21 Oct 2026The 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 | not stated | Source | Trend |
|---|---|---|---|
| GPQA Diamond | 56.1%#200 | Epoch AI Benchmarking Hub | |
| Arena Hard Prompts | 1278#277 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 57.5%#392 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 3.8%#568 | Artificial Analysis | |
| CritPt | 0.0%#315 | Artificial Analysis |
Coding
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Coding | 1306#275 | Arena (LMArena) |
Agents & tools
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Terminal-Bench Hard | 3.8%#298 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 0.0%#393 | Artificial Analysis | |
| BFCL Overall | 28.8%#52 | Berkeley Function Calling Leaderboard | |
| BALROGnot in index | 11.6%#33 | BALROG |
Maths
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| OTIS Mock AIME | 13.8%#227 | Epoch AI Benchmarking Hub | |
| MATH Level 5 | 64.9%#42 | Epoch AI Benchmarking Hub | |
| Arena Maths | 1265#259 | Arena (LMArena) |
Knowledge
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-Omniscience | -55.7#439 | Artificial Analysis | |
| MMLUnot in index | 84.8%#9 | Epoch AI Benchmarking Hub | |
| AA-Omniscience: accuracynot in index | 14.1%#444 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 18.8%#293 | Artificial Analysis |
Instruction following
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| IFBench | 23.5%#399 | Artificial Analysis | |
| Arena Instruction Followingnot in index | 1245#282 | Arena (LMArena) |
Human preference
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Text | 1256#287 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1210#286 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1242#282 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1267#281 | Arena (LMArena) | |
| Lech Mazur Creative Writingnot in index | 62.6%#38 | Lech Mazur |
Long context
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-LCR | 0.0%#493 | Artificial Analysis |
Composite
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Epoch Capabilities Indexnot in index | 130.5#143 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 5.9#551 | Artificial Analysis |
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 |
|---|---|---|---|---|---|---|
| DeepInfra | 80 tok/s | 0.37 s | $0.070 | $0.140 | 16k | bf16 |
Price history
The price above, per 1M tokens, every day it changed.
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. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0002 | – | 10.6 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0019 | – | 18.5 s |
| Code edit | 6,000 / 1,500 | $0.0015 | – | 42.2 s |
| Agentic coding session | 60,000 / 4,000 | $0.0098 | – | 1.8 min |
| Structured extraction | 2,000 / 200 | $0.0004 | – | 7.9 s |
More from Microsoft
See also
Questions people ask
- What are Phi-4's benchmark scores?
- As of 25 Oct 2026, Phi-4 scores 38.9 ±3.8 on the BenchLeader Index, strongest in maths (42) and weakest in long context (23). That rests on 14 independent benchmark results, and every score on this page links to its publisher.
- How does Phi-4 rank?
- Phi-4 is #699 of 785 ranked configurations on 25 Oct 2026. 224 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 Phi-4 cost per 1M tokens?
- As of 25 Oct 2026, Phi-4 costs $0.130 per 1M input tokens and $0.500 per 1M output tokens, $0.223 blended (three input tokens per output token). The list price last changed on 9 Sept 2026.
- When was Phi-4 released?
- Microsoft released Phi-4 on 12 Dec 2024.
Data as of 25 Oct 2026. Compare with another model.
Cite as: BenchLeader, “Phi-4: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/phi-4.