Qwen3.5 397B-A17B vs o3-pro vs Kimi K2.5
Too close to call on our weighted score (Kimi K2.5 63, Qwen3.5 397B-A17B 62, o3-pro 49). The right pick depends on what you value most.Too close to call on our weighted score (Kimi K2.5 71, Qwen3.5 397B-A17B 70, o3-pro 69). The right pick depends on what you value most.
Alibaba (Qwen)
Qwen3.5 397B-A17B
62/100- ECI148.0
- Price$0.60 / $3.60
- Context262K
OpenAI
o3-pro
49/100- ECI147.5
- Price$20.00 / $80.00
- Context200K
Moonshot AI
Kimi K2.5
63/100- ECI148.8
- Price$0.60 / $3.00
- Context262K
Alibaba (Qwen)
Qwen3.5 397B-A17B
70/100- ECI148.0
- Price$0.60 / $3.60
- Context262K
OpenAI
o3-pro
69/100- ECI147.5
- Price$20.00 / $80.00
- Context200K
Moonshot AI
Kimi K2.5
71/100- ECI148.8
- Price$0.60 / $3.00
- Context262K
Too close to call
It is close. Our weighted score puts them within 1 points (Kimi K2.5 63/100, Qwen3.5 397B-A17B 62/100, o3-pro 49/100), so choose by what matters most for your work: Kimi K2.5 for raw capability. The score weighs capability 59%, price 29%, context window 12%.It is close. Our weighted score puts them within a point (Kimi K2.5 71/100, Qwen3.5 397B-A17B 70/100, o3-pro 69/100), so choose by what matters most for your work: Kimi K2.5 for raw capability. The score weighs capability 83%, context window 17%.
- CapabilityKimi K2.5Capabilities Index (ECI): Kimi K2.5 148.8 · Qwen3.5 397B-A17B 148.0 · o3-pro 147.5
- Lowest priceKimi K2.5Kimi K2.5 $1.20 · Qwen3.5 397B-A17B $1.35 · o3-pro $35.00 per 1M tokens (3:1 blend)
- Longest contextQwen3.5 397B-A17B and Kimi K2.5Qwen3.5 397B-A17B 262,144 · Kimi K2.5 262,144 · o3-pro 200,000 tokens
- Widest inputsQwen3.5 397B-A17B and Kimi K2.5Qwen3.5 397B-A17B: Text, Images, Video · o3-pro: Text, Images · Kimi K2.5: Text, Images, Video
- Self-hostingQwen3.5 397B-A17B and Kimi K2.5Publishes downloadable weights
| Measure | Weight | Qwen3.5 397B-A17B | o3-pro | Kimi K2.5 |
|---|---|---|---|---|
| CapabilityCapabilities Index (ECI) | 59% | 76 | 75 | 77 |
| Price | 29% | 42 | 0 | 44 |
| Context window | 12% | 43 | 39 | 43 |
| Overall | 100% | 62/100 | 49/100 | 63/100 |
Left out because at least one model lacks the data: inputs & features. The remaining weights were rescaled.
| Measure | Weight | Qwen3.5 397B-A17B | o3-pro | Kimi K2.5 |
|---|---|---|---|---|
| CapabilityCapabilities Index (ECI) | 83% | 76 | 75 | 77 |
| Context window | 17% | 43 | 39 | 43 |
| Overall | 100% | 70/100 | 69/100 | 71/100 |
Left out because at least one model lacks the data: inputs & features. The remaining weights were rescaled.
Every spec in one table
Highlighted cells lead their row. Unavailable specifications are marked “Not published”. Publisher results include their evaluation settings and source.
| Specification | |||
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | 148.0 | 147.5 | 148.8 (best) |
| ECI rank | #76 of 159 | #80 of 159 | #74 of 159 (best) |
| ECI confidence interval | 145.9–150.1 | 145.7–149.4 | 147.3–150.3 |
| ECI evaluation date | Feb 13, 2026 | Jun 10, 2025 | Jan 27, 2026 |
| Documented capabilities | Large open Qwen multimodal MoE for visual agents and long technical tasks | High-effort o3 tier for difficult technical reasoning and careful answers | Earlier Kimi frontier model for long-context agents, coding, and multimodal work |
| Model parameters | Not published | Not published | Not published |
| Output modalities | Text | Text | Text |
| GPQA DiamondGraduate-level science questions | 86.4%none | Not published | 87.6% (best) |
| FrontierMath Tiers 1–3Research-level mathematics | 31.2%none | Not published | Not published |
| OTIS Mock AIME 2024–2025Competition mathematics | 88.9% | Not published | 92.2% (best) |
| SWE-bench VerifiedFixing real GitHub issues | Not published | Not published | 73.8% |
| SimpleQA VerifiedShort factual questions | Not published | Not published | 34.3% |
| Publisher-reported benchmarks | |||
| SWE-Bench VerifiedPublisher-reported result | 76.4resolvedReported result | Not published | 70.8resolvedReported result |
| Aider PolyglotPublisher-reported result | Not published | 84.9percent correctReported resultJun 28, 2025 | Not published |
| SWE-Atlas Codebase QnAPublisher-reported result | Not published | Not published | 13.1scoreReported result |
| SWE-Atlas RefactoringPublisher-reported result | Not published | Not published | 20.95scoreReported result |
| SWE-Atlas Test WritingPublisher-reported result | Not published | Not published | 25.77scoreReported result |
| Price per million tokens | |||
| Input | $0.60 (best) | $20.00 | $0.60 (best) |
| Output | $3.60 | $80.00 | $3.00 (best) |
| Cached input | Not published | Not published | Not published |
| Blended (3:1) | $1.35 | $35.00 | $1.20 (best) |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Official Alibaba API | Official OpenAI API | Median of 36 providers |
| Limits | |||
| Context window | 262,144 tokens (best) | 200,000 tokens | 262,144 tokens (best) |
| Max input | Not published | Not published | Not published |
| Max output | 65,536 tokens | 100,000 tokens | 262,144 tokens (best) |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | Yes | Yes | Yes |
| PDFs | No | No | No |
| Audio | No | No | No |
| Video | Yes | No | Yes |
| Reasoning | Yes | Yeslow · medium · high | Yes |
| Tool calling | Yes | Yes | Yes |
| Structured output | Yes | Yes | Not published |
| File attachments | Yes | Yes | Yes |
| Temperature control | Yes | No | No |
| Reasoning control | budget | effort | always |
| Availability | |||
| Weights | Open | Proprietary | Open |
| API model ID | qwen3.5-397b-a17b | o3-pro | Not published |
| API providers | 36 (best) | 10 | 36 (best) |
| Released | Feb 15, 2026 | Jun 10, 2025 | Jan 27, 2026 |
| Knowledge cutoff | Not published | May 2024 | Jan 2025 |
| Publisher documentation | Publisher documentation · Model weights | Publisher documentation | Model weights |
| Source metadata updated | Feb 15, 2026 | Jun 10, 2025 | Jan 27, 2026 |
| Data sources | models.dev · Epoch AI | models.dev · Epoch AI | models.dev · Epoch AI |
| Catalog checked | Nov 22, 2026 | Nov 22, 2026 | Nov 22, 2026 |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
Qwen3.5 397B-A17B$13.20
o3-pro$360.00
Kimi K2.5$12.00
Which should you choose?
Which is better: Qwen3.5 397B-A17B, o3-pro or Kimi K2.5?
It is close. Our weighted score puts them within 1 points (Kimi K2.5 63/100, Qwen3.5 397B-A17B 62/100, o3-pro 49/100), so choose by what matters most for your work: Kimi K2.5 for raw capability. The score weighs capability 59%, price 29%, context window 12%.It is close. Our weighted score puts them within a point (Kimi K2.5 71/100, Qwen3.5 397B-A17B 70/100, o3-pro 69/100), so choose by what matters most for your work: Kimi K2.5 for raw capability. The score weighs capability 83%, context window 17%.
Which is cheaper, Qwen3.5 397B-A17B, o3-pro or Kimi K2.5?
Kimi K2.5 is cheaper at $0.60 input / $3.00 output per million tokens (median across 36 API providers). Qwen3.5 397B-A17B costs $0.60 input / $3.60 output per million tokens (official Alibaba API price); o3-pro costs $20.00 input / $80.00 output per million tokens (official OpenAI API price). At a typical mix of three input tokens to one output token, that is $1.20 per million tokens for Kimi K2.5 versus $1.35 for Qwen3.5 397B-A17B (1.1× as much) and $35.00 for o3-pro (29× as much).
Which scores higher on benchmarks?
Kimi K2.5 scores higher on the Capabilities Index (ECI): Kimi K2.5 148.8 (#74 of 159), Qwen3.5 397B-A17B 148.0 (#76 of 159) and o3-pro 147.5 (#80 of 159). The confidence ranges of the top two overlap (147.3–150.3 vs 145.9–150.1), so treat the gap as small.
Which is better for coding?
There are no published SWE-bench Verified results for Qwen3.5 397B-A17B and o3-pro yet, so there is no like-for-like coding score. On overall capability, Kimi K2.5 leads, which tends to carry over to coding, but test on your own codebase. All three support tool calling for agent workflows.
Which has the bigger context window?
Qwen3.5 397B-A17B and Kimi K2.5 have the largest context windows (262,144 and 262,144 tokens), against 200,000 for o3-pro. Maximum output per response: Qwen3.5 397B-A17B up to 65,536, o3-pro up to 100,000, Kimi K2.5 up to 262,144 tokens.
Which can read images, PDFs, audio or video?
Qwen3.5 397B-A17B accepts text, images and video; o3-pro accepts text and images; Kimi K2.5 accepts text, images and video. Qwen3.5 397B-A17B handles the widest range of inputs.
Are any of these open source?
Qwen3.5 397B-A17B and Kimi K2.5 publishes its weights and can be self-hosted; o3-pro is proprietary.
Which is newer?
Qwen3.5 397B-A17B is the newest, released Feb 15, 2026. Kimi K2.5 came out Jan 27, 2026; o3-pro came out Jun 10, 2025. Knowledge cutoff: o3-pro May 2024, Kimi K2.5 Jan 2025.
How do you decide the winner?
This rating uses capability (59%, independent benchmark results); price (29%, blended price per million tokens (3 input : 1 output), log scale); context window (12%, 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 22, 2026.This rating uses capability (83%, independent benchmark results); context window (17%, 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 22, 2026.