Llama 3.3 Nemotron Super 49B v1 vs Phi-4-mini vs Voxtral Small 24B 2507
Too close to call on our weighted score (Llama 3.3 Nemotron Super 49B v1 60, Phi-4-mini 58, Voxtral Small 24B 2507 55). The right pick depends on what you value most.
NVIDIALlama 3.3 Nemotron Super 49B v1
60/100- ECI—
- Price$0.15 / $0.15
- Context131K
MicrosoftPhi-4-mini
58/100- ECI—
- Price$0.075 / $0.30
- Context128K
Mistral AIVoxtral Small 24B 2507
55/100- ECI—
- Price$0.10 / $0.30
- Context33K
Too close to call
It is close. Our weighted score puts them within 2 points (Llama 3.3 Nemotron Super 49B v1 60/100, Phi-4-mini 58/100, Voxtral Small 24B 2507 55/100), so choose by what matters most for your work: Phi-4-mini on price and Llama 3.3 Nemotron Super 49B v1 for long inputs. 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 pricePhi-4-miniPhi-4-mini $0.131 · Llama 3.3 Nemotron Super 49B v1 $0.15 · Voxtral Small 24B 2507 $0.15 per 1M tokens (3:1 blend)
- Longest contextLlama 3.3 Nemotron Super 49B v1Llama 3.3 Nemotron Super 49B v1 131,072 · Phi-4-mini 128,000 · Voxtral Small 24B 2507 32,768 tokens
- Widest inputsVoxtral Small 24B 2507Llama 3.3 Nemotron Super 49B v1: Text · Phi-4-mini: Text · Voxtral Small 24B 2507: Text, Audio
- Self-hostingAll open weightsEvery model here can be downloaded and run on your own hardware
| Measure | Weight | Llama 3.3 Nemotron Super 49B v1 | Phi-4-mini | Voxtral Small 24B 2507 |
|---|---|---|---|---|
| Price | 50% | 89 | 92 | 89 |
| Inputs & features | 30% | 35 | 25 | 35 |
| Context window | 20% | 24 | 24 | 0 |
| Overall | 100% | 60/100 | 58/100 | 55/100 |
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 | Llama 3.3 Nemotron Super 49B v1NVIDIA | Phi-4-miniMicrosoft | Voxtral Small 24B 2507Mistral AI |
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | — | — | — |
| ECI rank | — | — | — |
| Price per million tokens | |||
| Input | $0.15 | $0.075 (best) | $0.10 |
| Output | $0.15 (best) | $0.30 | $0.30 |
| Cached input | — | — | — |
| Blended (3:1) | $0.15 | $0.131 (best) | $0.15 |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Median of 1 providers | Official Azure API | Official Mistral API |
| Limits | |||
| Context window | 131,072 tokens (best) | 128,000 tokens | 32,768 tokens |
| Max output | 131,072 tokens (best) | 4,096 tokens | 32,768 tokens |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | No | No | No |
| PDFs | No | No | No |
| Audio | No | No | Yes |
| Video | No | No | No |
| Reasoning | Yes | No | No |
| Tool calling | Yes | Yes | Yes |
| Structured output | No | No | No |
| Availability | |||
| Weights | Open | Open | OpenApache 2.0 |
| API model ID | nvidia/llama-3.3-nemotron-super-49b-v1 | phi-4-mini | voxtral-small-latest |
| API providers | 2 | 1 | 7 (best) |
| Released | Apr 7, 2025 | Dec 11, 2024 | Jul 15, 2025 |
| Knowledge cutoff | — | Oct 2023 | — |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
- Llama 3.3 Nemotron Super 49B v1$1.80
- Phi-4-mini$1.35
- Voxtral Small 24B 2507$1.60
Which should you choose?
Which is better: Llama 3.3 Nemotron Super 49B v1, Phi-4-mini or Voxtral Small 24B 2507?
It is close. Our weighted score puts them within 2 points (Llama 3.3 Nemotron Super 49B v1 60/100, Phi-4-mini 58/100, Voxtral Small 24B 2507 55/100), so choose by what matters most for your work: Phi-4-mini on price and Llama 3.3 Nemotron Super 49B v1 for long inputs. 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, Llama 3.3 Nemotron Super 49B v1, Phi-4-mini or Voxtral Small 24B 2507?
Phi-4-mini is cheaper at $0.075 input / $0.30 output per million tokens (official Azure API price). Llama 3.3 Nemotron Super 49B v1 costs $0.15 input / $0.15 output per million tokens (median across 1 API provider; free on Nvidia); Voxtral Small 24B 2507 costs $0.10 input / $0.30 output per million tokens (official Mistral API price). At a typical mix of three input tokens to one output token, that is $0.131 per million tokens for Phi-4-mini versus $0.15 for Llama 3.3 Nemotron Super 49B v1 (1.1× as much) and $0.15 for Voxtral Small 24B 2507 (1.1× as much).
Which scores higher on benchmarks?
There is no independent benchmark that covers all three models yet. Llama 3.3 Nemotron Super 49B v1 has not been scored yet, Phi-4-mini has not been scored yet and Voxtral Small 24B 2507 has not been scored yet.
Which is better for coding?
There are no published SWE-bench Verified results for Llama 3.3 Nemotron Super 49B v1, Phi-4-mini and Voxtral Small 24B 2507 yet, so there is no like-for-like coding score. Test both on a sample of your own tasks. All three support tool calling for agent workflows.
Which has the bigger context window?
Llama 3.3 Nemotron Super 49B v1 has the largest context window at 131,072 tokens, against 128,000 for Phi-4-mini and 32,768 for Voxtral Small 24B 2507. Maximum output per response: Llama 3.3 Nemotron Super 49B v1 up to 131,072, Phi-4-mini up to 4,096, Voxtral Small 24B 2507 up to 32,768 tokens.
Which can read images, PDFs, audio or video?
Llama 3.3 Nemotron Super 49B v1 accepts text; Phi-4-mini accepts text; Voxtral Small 24B 2507 accepts text and audio. Voxtral Small 24B 2507 handles the widest range of inputs.
Are any of these open source?
Yes, all three publish their weights (Apache 2.0), so you can self-host them.
Which is newer?
Voxtral Small 24B 2507 is the newest, released Jul 15, 2025. Llama 3.3 Nemotron Super 49B v1 came out Apr 7, 2025; Phi-4-mini came out Dec 11, 2024. Knowledge cutoff: Phi-4-mini Oct 2023.
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 Oct 4, 2026.
