Codestral vs Qwen2.5-VL 7B Instruct vs Sonar Pro
Qwen2.5-VL 7B Instruct comes out ahead, 51 to 48 and 20 on our weighted score, though Codestral is 14% cheaper per token.
Mistral AICodestral
48/100- ECI—
- Price$0.30 / $0.90
- Context256K
- Our pick
Alibaba (Qwen)Qwen2.5-VL 7B Instruct
51/100- ECI—
- Price$0.35 / $1.05
- Context131K
PerplexitySonar Pro
20/100- ECI—
- Price$3.00 / $15.00
- Context200K
Qwen2.5-VL 7B Instruct is our pick
Qwen2.5-VL 7B Instruct is the better all-round choice, scoring 51/100 against Codestral (48) and Sonar Pro (20). It leads on inputs & features. Codestral wins on price and context window. 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 priceCodestralCodestral $0.45 · Qwen2.5-VL 7B Instruct $0.525 · Sonar Pro $6.00 per 1M tokens (3:1 blend)
- Longest contextCodestralCodestral 256,000 · Sonar Pro 200,000 · Qwen2.5-VL 7B Instruct 131,072 tokens
- Widest inputsQwen2.5-VL 7B Instruct and Sonar ProCodestral: Text · Qwen2.5-VL 7B Instruct: Text, Images · Sonar Pro: Text, Images
- Self-hostingCodestral and Qwen2.5-VL 7B InstructPublishes downloadable weights
| Measure | Weight | Codestral | Qwen2.5-VL 7B Instruct | Sonar Pro |
|---|---|---|---|---|
| Price | 50% | 66 | 63 | 13 |
| Inputs & features | 30% | 25 | 50 | 25 |
| Context window | 20% | 36 | 24 | 32 |
| Overall | 100% | 48/100 | 51/100 | 20/100 |
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 | CodestralMistral AI | Qwen2.5-VL 7B InstructAlibaba (Qwen) | Sonar ProPerplexity |
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | — | — | — |
| ECI rank | — | — | — |
| Price per million tokens | |||
| Input | $0.30 (best) | $0.35 | $3.00 |
| Output | $0.90 (best) | $1.05 | $15.00 |
| Cached input | $0.03 | — | — |
| Blended (3:1) | $0.45 (best) | $0.525 | $6.00 |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Official Mistral API | Official Alibaba API | Official Perplexity API |
| Limits | |||
| Context window | 256,000 tokens (best) | 131,072 tokens | 200,000 tokens |
| Max output | 4,096 tokens | 8,192 tokens (best) | 8,192 tokens (best) |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | No | Yes | Yes |
| PDFs | No | No | No |
| Audio | No | No | No |
| Video | No | No | No |
| Reasoning | No | No | No |
| Tool calling | Yes | Yes | No |
| Structured output | No | No | No |
| Availability | |||
| Weights | Open | Open | Proprietary |
| API model ID | codestral-latest | qwen2-5-vl-7b-instruct | sonar-pro |
| API providers | 3 | 1 | 5 (best) |
| Released | May 29, 2024 | Sep 2024 | Jan 1, 2024 |
| Knowledge cutoff | Oct 2024 | Apr 2024 | Sep 1, 2025 |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
- Codestral$4.80
- Qwen2.5-VL 7B Instruct$5.60
- Sonar Pro$60.00
Which should you choose?
Which is better: Codestral, Qwen2.5-VL 7B Instruct or Sonar Pro?
Qwen2.5-VL 7B Instruct is the better all-round choice, scoring 51/100 against Codestral (48) and Sonar Pro (20). It leads on inputs & features. Codestral wins on price and context window. 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, Codestral, Qwen2.5-VL 7B Instruct or Sonar Pro?
Codestral is cheaper at $0.30 input / $0.90 output per million tokens (official Mistral API price). Qwen2.5-VL 7B Instruct costs $0.35 input / $1.05 output per million tokens (official Alibaba API price); Sonar Pro costs $3.00 input / $15.00 output per million tokens (official Perplexity API price). At a typical mix of three input tokens to one output token, that is $0.45 per million tokens for Codestral versus $0.525 for Qwen2.5-VL 7B Instruct (1.2× as much) and $6.00 for Sonar Pro (13× as much).
Which scores higher on benchmarks?
There is no independent benchmark that covers all three models yet. Codestral has not been scored yet, Qwen2.5-VL 7B Instruct has not been scored yet and Sonar Pro has not been scored yet.
Which is better for coding?
There are no published SWE-bench Verified results for Codestral, Qwen2.5-VL 7B Instruct and Sonar Pro yet, so there is no like-for-like coding score. Test both on a sample of your own tasks. Note that Sonar Pro does not support tool calling, which most coding agents need.
Which has the bigger context window?
Codestral has the largest context window at 256,000 tokens, against 200,000 for Sonar Pro and 131,072 for Qwen2.5-VL 7B Instruct. Maximum output per response: Codestral up to 4,096, Qwen2.5-VL 7B Instruct up to 8,192, Sonar Pro up to 8,192 tokens.
Which can read images, PDFs, audio or video?
Codestral accepts text; Qwen2.5-VL 7B Instruct accepts text and images; Sonar Pro accepts text and images. Qwen2.5-VL 7B Instruct handles the widest range of inputs.
Are any of these open source?
Codestral and Qwen2.5-VL 7B Instruct publishes its weights and can be self-hosted; Sonar Pro is proprietary.
Which is newer?
Qwen2.5-VL 7B Instruct is the newest, released Sep 2024. Codestral came out May 29, 2024; Sonar Pro came out Jan 1, 2024. Knowledge cutoff: Codestral Oct 2024, Qwen2.5-VL 7B Instruct Apr 2024, Sonar Pro Sep 1, 2025.
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.
