Compare / Phi 4 vs Mistral Medium 3.5
Phi 4 vs Mistral Medium 3.5
Which is cheaper, which is more capable, and which does better at each kind of work, from 18 results on 3 sources that measured both models.
Phi 4 or Mistral Medium 3.5? The short answer
- Phi 4 is cheaper: $0.088 against $3.00 per million tokens, 34 times less.
- Mistral Medium 3.5 scores higher on the Artificial Analysis Intelligence Index: 14.2 against 5.9.
- Mistral Medium 3.5 writes faster: 163 tokens a second against 42.
- On our business benchmarks, Phi 4 ranks higher for sticking to the facts.
- On our business benchmarks, Mistral Medium 3.5 ranks higher for agents and tool use, reasoning and knowledge and writing.
At a glance
| Phi 4 | Mistral Medium 3.5 | |
|---|---|---|
| Price per million tokens | $0.070 in · $0.14 out | $1.50 in · $7.50 out |
| Blended (3 in : 1 out) | $0.088 | $3.00 |
| Cheapest host, blended | – | $3.00 (Mistral) |
| Intelligence Index | 5.9 | 14.2 |
| Spring Prompt overall | – | 21 |
| Output speed, tokens a second | 42 | 163 |
| Context window | 16,384 tokens | 262,144 tokens |
| Weights | Open | Closed |
| Developer based in | the US | France |
Prices are the developer's list price where we have read it, otherwise the typical host on OpenRouter; see every model's price. Intelligence and speed from Artificial Analysis.
Which is better for what
Each model's rank among models available today, for each kind of work, from our own benchmarks and licensed sources. The better rank is in green.
| Use case | Phi 4 | Mistral Medium 3.5 |
|---|---|---|
| Product listingsTurning a sparse product feed and photos into listings that can go live | not measured | 13 of 19 |
| Decks from an analysisTurning a finished analysis into a deck you could present as it is | not measured | 14 of 18 |
| Marketing planningPlanning a year of ad spend without overspending | not measured | 18 of 18 |
| Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls | 122 of 161 | 87 of 161 |
| Reasoning and knowledgeHard questions across science, maths and general knowledge | 178 of 195 | 119 of 195 |
| WritingWhat people prefer in blind comparisons, and judged writing quality | 175 of 187 | 85 of 187 |
| Sticking to the factsSummarising without inventing things, and factual answers | 3 of 153 | 99 of 153 |
| Speed under pressureGood decisions against a real clock (fast chess) | not measured | 8 of 21 |
From our research
Every shared result
| Benchmark | Metric | Phi 4 | Mistral Medium 3.5 |
|---|---|---|---|
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,256 | 1,430 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,210 | 1,398 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,268 | 1,434 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,245 | 1,421 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,256 | 1,426 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,229 | 1,404 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 5.90 | 14.2 |
| Artificial Analysisreported | GPQA Diamond% of questions | 57.5% | 74.8% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 3.8% | 13.8% |
| Artificial Analysisreported | IFBench% of instructions | 23.5% | 68.8% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 0.0% | 69.3% |
| Artificial Analysisreported | Output speedtokens per second | 42.3 | 163 |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 3.8% | 33.3% |
| Artificial Analysisreported | Time to first answer tokenseconds | 1.03 | 12.9 |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 0.0% | 94.2% |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 19.0% | 18.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.34 | 0.35 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 25.7 | 45.5 |
Bold green marks the better value on that metric. Where a source reports ranges that overlap, the difference may not be meaningful; see the benchmark page for ranges.
Quick answers
Is Phi 4 better than Mistral Medium 3.5?
It depends on the task. Among models available today, on our benchmarks and licensed sources, Phi 4 ranks higher for sticking to the facts; Mistral Medium 3.5 ranks higher for agents and tool use, reasoning and knowledge and writing.
Which is cheaper, Phi 4 or Mistral Medium 3.5?
Phi 4 costs $0.088 per million tokens (three input to one output), against $3.00 for Mistral Medium 3.5.
Which is faster, Phi 4 or Mistral Medium 3.5?
Mistral Medium 3.5 writes about 163 tokens a second on its usual API, against 42 for Phi 4 (Artificial Analysis).
Run Phi 4 and Mistral Medium 3.5 on your own prompt
Benchmarks aren't your data. Try both side by side in the playground, with the cost of every answer.