Vision Small vs Qwen2.5-VL 72B Instruct vs Qwen Max
Vision Small comes out ahead, 61 to 44 and 24 on our weighted score, though Qwen2.5-VL 72B Instruct is 41% cheaper per token.
- Our pick
VisparkVision Small
61/100- ECI—
- Price$1.05 / $3.16
- Context1M
Alibaba (Qwen)Qwen2.5-VL 72B Instruct
44/100- ECI—
- Price$0.905 / $1.00
- Context131K
- Alibaba (Qwen)
Qwen Max
24/100- ECI—
- Price$1.60 / $6.40
- Context33K
Vision Small is our pick
Vision Small is the better all-round choice, scoring 61/100 against Qwen2.5-VL 72B Instruct (44) and Qwen Max (24). It leads on inputs & features and context window. Qwen2.5-VL 72B Instruct 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 priceQwen2.5-VL 72B InstructQwen2.5-VL 72B Instruct $0.93 · Vision Small $1.58 · Qwen Max $2.80 per 1M tokens (3:1 blend)
- Longest contextVision SmallVision Small 1,000,000 · Qwen2.5-VL 72B Instruct 131,072 · Qwen Max 32,768 tokens
- Widest inputsVision SmallVision Small: Text, Images, PDFs, Audio, Video · Qwen2.5-VL 72B Instruct: Text, Images · Qwen Max: Text
- Self-hostingQwen2.5-VL 72B InstructPublishes downloadable weights
| Measure | Weight | Vision Small | Qwen2.5-VL 72B Instruct | Qwen Max |
|---|---|---|---|---|
| Price | 50% | 39 | 49 | 27 |
| Inputs & features | 30% | 100 | 50 | 35 |
| Context window | 20% | 60 | 25 | 0 |
| Overall | 100% | 61/100 | 44/100 | 24/100 |
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 | Vision SmallVispark | Qwen2.5-VL 72B InstructAlibaba (Qwen) | Qwen MaxAlibaba (Qwen) |
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | — | — | — |
| ECI rank | — | — | — |
| Price per million tokens | |||
| Input | $1.05 | $0.905 (best) | $1.60 |
| Output | $3.16 | $1.00 (best) | $6.40 |
| Cached input | — | — | — |
| Blended (3:1) | $1.58 | $0.93 (best) | $2.80 |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Official Vispark API | Median of 6 providers | Official Alibaba API |
| Limits | |||
| Context window | 1,000,000 tokens (best) | 131,072 tokens | 32,768 tokens |
| Max output | 65,536 tokens (best) | 8,192 tokens | 8,192 tokens |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | Yes | Yes | No |
| PDFs | Yes | No | No |
| Audio | Yes | No | No |
| Video | Yes | No | No |
| Reasoning | Yes | No | No |
| Tool calling | Yes | Yes | Yes |
| Structured output | Yes | No | Yes |
| Availability | |||
| Weights | Proprietary | Open | Proprietary |
| API model ID | vispark/vision-small | — | qwen-max |
| API providers | 1 | 6 | 8 (best) |
| Released | May 15, 2024 | Jan 26, 2025 | Apr 3, 2024 |
| Knowledge cutoff | — | Apr 2024 | Apr 2024 |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
- Vision Small$16.82
- Qwen2.5-VL 72B Instruct$11.06
- Qwen Max$28.80
Which should you choose?
Which is better: Vision Small, Qwen2.5-VL 72B Instruct or Qwen Max?
Vision Small is the better all-round choice, scoring 61/100 against Qwen2.5-VL 72B Instruct (44) and Qwen Max (24). It leads on inputs & features and context window. Qwen2.5-VL 72B Instruct 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, Vision Small, Qwen2.5-VL 72B Instruct or Qwen Max?
Qwen2.5-VL 72B Instruct is cheaper at $0.905 input / $1.00 output per million tokens (median across 6 API providers). Vision Small costs $1.05 input / $3.16 output per million tokens (official Vispark API price); Qwen Max costs $1.60 input / $6.40 output per million tokens (official Alibaba API price). At a typical mix of three input tokens to one output token, that is $0.93 per million tokens for Qwen2.5-VL 72B Instruct versus $1.58 for Vision Small (1.7× as much) and $2.80 for Qwen Max (3× as much).
Which scores higher on benchmarks?
There is no independent benchmark that covers all three models yet. Vision Small has not been scored yet, Qwen2.5-VL 72B Instruct has not been scored yet and Qwen Max has not been scored yet.
Which is better for coding?
There are no published SWE-bench Verified results for Vision Small, Qwen2.5-VL 72B Instruct and Qwen Max 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?
Vision Small has the largest context window at 1,000,000 tokens, against 131,072 for Qwen2.5-VL 72B Instruct and 32,768 for Qwen Max. Maximum output per response: Vision Small up to 65,536, Qwen2.5-VL 72B Instruct up to 8,192, Qwen Max up to 8,192 tokens.
Which can read images, PDFs, audio or video?
Vision Small accepts text, images, PDFs, audio and video; Qwen2.5-VL 72B Instruct accepts text and images; Qwen Max accepts text. Vision Small handles the widest range of inputs.
Are any of these open source?
Qwen2.5-VL 72B Instruct publishes its weights and can be self-hosted; Vision Small and Qwen Max is proprietary.
Which is newer?
Qwen2.5-VL 72B Instruct is the newest, released Jan 26, 2025. Vision Small came out May 15, 2024; Qwen Max came out Apr 3, 2024. Knowledge cutoff: Qwen2.5-VL 72B Instruct Apr 2024, Qwen Max Apr 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.
