Mistral Small 3 vs Phi-4-multimodal-instruct vs Gemma 3 27B IT
Phi-4-multimodal-instruct comes out ahead, 72 to 62 and 50 on our weighted score, though Mistral Small 3 is 28% cheaper per token.
Mistral AIMistral Small 3
50/100- ECI—
- Price$0.05 / $0.08
- Context33K
- Our pick
MicrosoftPhi-4-multimodal-instruct
72/100- ECI—
- Price$0.07 / $0.11
- Context131K
GoogleGemma 3 27B IT
62/100- ECI129.9
- Price$0.11 / $0.30
- Context131K
Phi-4-multimodal-instruct is our pick
Phi-4-multimodal-instruct is the better all-round choice, scoring 72/100 against Gemma 3 27B IT (62) and Mistral Small 3 (50). It leads on inputs & features. Mistral Small 3 wins on price. The score weighs price 50%, inputs & features 30%, context window 20%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.
- CapabilityNot enough dataNo independent benchmark covers every model here yet
- Lowest priceMistral Small 3Mistral Small 3 $0.058 · Phi-4-multimodal-instruct $0.08 · Gemma 3 27B IT $0.157 per 1M tokens (3:1 blend)
- Longest contextPhi-4-multimodal-instruct and Gemma 3 27B ITPhi-4-multimodal-instruct 131,072 · Gemma 3 27B IT 131,072 · Mistral Small 3 32,768 tokens
- Widest inputsPhi-4-multimodal-instructMistral Small 3: Text · Phi-4-multimodal-instruct: Text, Images, Audio · Gemma 3 27B IT: Text, Images
- Self-hostingAll open weightsEvery model here can be downloaded and run on your own hardware
| Measure | Weight | Mistral Small 3 | Phi-4-multimodal-instruct | Gemma 3 27B IT |
|---|---|---|---|---|
| Price | 50% | 100 | 98 | 84 |
| Inputs & features | 30% | 0 | 60 | 50 |
| Context window | 20% | 0 | 25 | 25 |
| Overall | 100% | 50/100 | 72/100 | 62/100 |
Left out because at least one model lacks the data: capability. The remaining weights were rescaled.
Every spec in one table
Highlighted cells lead their row. Dashes mean the data is not published.
| Specification | Mistral Small 3Mistral AI | Phi-4-multimodal-instructMicrosoft | Gemma 3 27B ITGoogle |
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | — | — | 129.9 |
| ECI rank | — | — | #137 of 165 |
| GPQA DiamondGraduate-level science questions | 47.3% | — | 47.7% (best) |
| OTIS Mock AIME 2024–2025Competition mathematics | 6.7% | — | 22.5% (best) |
| Price per million tokens | |||
| Input | $0.05 (best) | $0.07 | $0.11 |
| Output | $0.08 (best) | $0.11 | $0.30 |
| Cached input | — | — | — |
| Blended (3:1) | $0.058 (best) | $0.08 | $0.157 |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Median of 6 providers | Median of 1 providers | Median of 16 providers |
| Limits | |||
| Context window | 32,768 tokens | 131,072 tokens (best) | 131,072 tokens (best) |
| Max output | 16,384 tokens | 131,072 tokens (best) | 131,072 tokens (best) |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | No | Yes | Yes |
| PDFs | No | No | No |
| Audio | No | Yes | No |
| Video | No | No | No |
| Reasoning | No | No | No |
| Tool calling | No | Yes | Yes |
| Structured output | No | No | No |
| Availability | |||
| Weights | Open | OpenMIT | Open |
| API model ID | — | — | — |
| API providers | 6 | 1 | 17 (best) |
| Released | Jan 30, 2025 | Feb 26, 2025 | Mar 12, 2025 |
| Knowledge cutoff | Oct 31, 2023 | Jun 2024 | Aug 2024 |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
- Mistral Small 3$0.66
- Phi-4-multimodal-instruct$0.92
- Gemma 3 27B IT$1.69
Which should you choose?
Which is better: Mistral Small 3, Phi-4-multimodal-instruct or Gemma 3 27B IT?
Phi-4-multimodal-instruct is the better all-round choice, scoring 72/100 against Gemma 3 27B IT (62) and Mistral Small 3 (50). It leads on inputs & features. Mistral Small 3 wins on price. The score weighs price 50%, inputs & features 30%, context window 20%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.
Which is cheaper, Mistral Small 3, Phi-4-multimodal-instruct or Gemma 3 27B IT?
Mistral Small 3 is cheaper at $0.05 input / $0.08 output per million tokens (median across 6 API providers). Phi-4-multimodal-instruct costs $0.07 input / $0.11 output per million tokens (median across 1 API provider); Gemma 3 27B IT costs $0.11 input / $0.30 output per million tokens (median across 16 API providers). At a typical mix of three input tokens to one output token, that is $0.058 per million tokens for Mistral Small 3 versus $0.08 for Phi-4-multimodal-instruct (1.4× as much) and $0.157 for Gemma 3 27B IT (2.7× as much).
Which scores higher on benchmarks?
There is no independent benchmark that covers all three models yet. Mistral Small 3 has not been scored yet, Phi-4-multimodal-instruct has not been scored yet and Gemma 3 27B IT has an ECI of 129.9.
Which is better for coding?
There are no published SWE-bench Verified results for Mistral Small 3, Phi-4-multimodal-instruct and Gemma 3 27B IT yet, so there is no like-for-like coding score. Test both on a sample of your own tasks. Note that Mistral Small 3 does not support tool calling, which most coding agents need.
Which has the bigger context window?
Phi-4-multimodal-instruct and Gemma 3 27B IT have the largest context windows (131,072 and 131,072 tokens), against 32,768 for Mistral Small 3. Maximum output per response: Mistral Small 3 up to 16,384, Phi-4-multimodal-instruct up to 131,072, Gemma 3 27B IT up to 131,072 tokens.
Which can read images, PDFs, audio or video?
Mistral Small 3 accepts text; Phi-4-multimodal-instruct accepts text, images and audio; Gemma 3 27B IT accepts text and images. Phi-4-multimodal-instruct handles the widest range of inputs.
Are any of these open source?
Yes, all three publish their weights (MIT), so you can self-host them.
Which is newer?
Gemma 3 27B IT is the newest, released Mar 12, 2025. Phi-4-multimodal-instruct came out Feb 26, 2025; Mistral Small 3 came out Jan 30, 2025. Knowledge cutoff: Mistral Small 3 Oct 31, 2023, Phi-4-multimodal-instruct Jun 2024, Gemma 3 27B IT Aug 2024.
How do you decide the winner?
Each model gets a 0–100 score on capability (50%, independent benchmark results); price (25%, blended price per million tokens (3 input : 1 output), log scale); inputs & features (15%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (10%, maximum tokens per request, log scale). Dimensions missing for any model are dropped and the remaining weights rescaled, so every model is judged on the same evidence. Specs and prices come from public model listings and the labs’ own API pages; capability scores come from independent benchmark runs. Data updated Nov 2, 2026.
