Qwen3 235B-A22B vs Kimi K2 Instruct vs DeepSeek R1 0528
Too close to call on our weighted score (DeepSeek R1 0528 55, Kimi K2 Instruct 53, Qwen3 235B-A22B 53). The right pick depends on what you value most.
Price in rating: Included · Exclude price
Alibaba (Qwen)
Qwen3 235B-A22B
53/100- ECI138.4
- Price$0.70 / $2.80
- Context131K
Moonshot AI
Kimi K2 Instruct
53/100- ECI141.3
- Price$0.57 / $2.30
- Context131K
DeepSeek
DeepSeek R1 0528
55/100- ECI140.8
- Price$0.512 / $2.19
- Context164K
Too close to call
It is close. Our weighted score puts them within 2 points (DeepSeek R1 0528 55/100, Kimi K2 Instruct 53/100, Qwen3 235B-A22B 53/100), so choose by what matters most for your work: Kimi K2 Instruct for raw capability, DeepSeek R1 0528 on price and DeepSeek R1 0528 for long inputs. The score weighs capability 50%, price 25%, inputs & features 15%, context window 10%.
- CapabilityKimi K2 InstructCapabilities Index (ECI): Kimi K2 Instruct 141.3 · DeepSeek R1 0528 140.8 · Qwen3 235B-A22B 138.4
- Lowest priceDeepSeek R1 0528DeepSeek R1 0528 $0.932 · Kimi K2 Instruct $1.00 · Qwen3 235B-A22B $1.23 per 1M tokens (3:1 blend)
- Longest contextDeepSeek R1 0528DeepSeek R1 0528 163,840 · Qwen3 235B-A22B 131,072 · Kimi K2 Instruct 131,072 tokens
- Widest inputsSame inputsQwen3 235B-A22B: Text · Kimi K2 Instruct: Text · DeepSeek R1 0528: Text
- Self-hostingAll open weightsEvery model here can be downloaded and run on your own hardware
| Measure | Weight | Qwen3 235B-A22B | Kimi K2 Instruct | DeepSeek R1 0528 |
|---|---|---|---|---|
| CapabilityCapabilities Index (ECI) | 50% | 66 | 69 | 69 |
| Price | 25% | 44 | 48 | 49 |
| Inputs & features | 15% | 35 | 25 | 35 |
| Context window | 10% | 32 | 32 | 36 |
| Overall | 100% | 53/100 | 53/100 | 55/100 |
Every spec in one table
Highlighted cells lead their row. Dashes mean the data is not published.
| Specification | |||
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | 138.4 | 141.3 (best) | 140.8 |
| ECI rank | #119 of 158 | #108 of 158 (best) | #113 of 158 |
| GPQA DiamondGraduate-level science questions | 70.7% | — | 76.3% (best) |
| OTIS Mock AIME 2024–2025Competition mathematics | — | — | 66.4% |
| Price per million tokens | |||
| Input | $0.70 | $0.57 | $0.512 (best) |
| Output | $2.80 | $2.30 | $2.19 (best) |
| Cached input | — | — | — |
| Blended (3:1) | $1.23 | $1.00 | $0.932 (best) |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Official Alibaba API | Median of 11 providers | Median of 16 providers |
| Limits | |||
| Context window | 131,072 tokens | 131,072 tokens | 163,840 tokens (best) |
| Max output | 16,384 tokens | 32,768 tokens (best) | 32,768 tokens (best) |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | No | No | No |
| PDFs | No | No | No |
| Audio | No | No | No |
| Video | No | No | No |
| Reasoning | Yes | No | Yes |
| Tool calling | Yes | Yes | Yes |
| Structured output | No | No | No |
| Availability | |||
| Weights | Open | OpenModified MIT | OpenMIT |
| API model ID | qwen3-235b-a22b | — | — |
| API providers | 7 | 11 | 16 (best) |
| Released | Apr 28, 2025 | Jul 11, 2025 | May 28, 2025 |
| Knowledge cutoff | Apr 2025 | — | — |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
Qwen3 235B-A22B$12.60
Kimi K2 Instruct$10.30
DeepSeek R1 0528$9.50
Which should you choose?
Which is better: Qwen3 235B-A22B, Kimi K2 Instruct or DeepSeek R1 0528?
It is close. Our weighted score puts them within 2 points (DeepSeek R1 0528 55/100, Kimi K2 Instruct 53/100, Qwen3 235B-A22B 53/100), so choose by what matters most for your work: Kimi K2 Instruct for raw capability, DeepSeek R1 0528 on price and DeepSeek R1 0528 for long inputs. The score weighs capability 50%, price 25%, inputs & features 15%, context window 10%.
Which is cheaper, Qwen3 235B-A22B, Kimi K2 Instruct or DeepSeek R1 0528?
DeepSeek R1 0528 is cheaper at $0.512 input / $2.19 output per million tokens (median across 16 API providers). Kimi K2 Instruct costs $0.57 input / $2.30 output per million tokens (median across 11 API providers); Qwen3 235B-A22B costs $0.70 input / $2.80 output per million tokens (official Alibaba API price). At a typical mix of three input tokens to one output token, that is $0.932 per million tokens for DeepSeek R1 0528 versus $1.00 for Kimi K2 Instruct (1.1× as much) and $1.23 for Qwen3 235B-A22B (1.3× as much).
Which scores higher on benchmarks?
Kimi K2 Instruct scores higher on the Capabilities Index (ECI): Kimi K2 Instruct 141.3 (#108 of 158), DeepSeek R1 0528 140.8 (#113 of 158) and Qwen3 235B-A22B 138.4 (#119 of 158). The confidence ranges of the top two overlap (138.3–144.0 vs 138.9–142.5), so treat the gap as small.
Which is better for coding?
There are no published SWE-bench Verified results for Qwen3 235B-A22B, Kimi K2 Instruct and DeepSeek R1 0528 yet, so there is no like-for-like coding score. On overall capability, Kimi K2 Instruct 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?
DeepSeek R1 0528 has the largest context window at 163,840 tokens, against 131,072 for Qwen3 235B-A22B and 131,072 for Kimi K2 Instruct. Maximum output per response: Qwen3 235B-A22B up to 16,384, Kimi K2 Instruct up to 32,768, DeepSeek R1 0528 up to 32,768 tokens.
Which can read images, PDFs, audio or video?
Qwen3 235B-A22B accepts text; Kimi K2 Instruct accepts text; DeepSeek R1 0528 accepts text. They handle the same number of input types.
Are any of these open source?
Yes, all three publish their weights (Modified MIT and MIT), so you can self-host them.
Which is newer?
Kimi K2 Instruct is the newest, released Jul 11, 2025. DeepSeek R1 0528 came out May 28, 2025; Qwen3 235B-A22B came out Apr 28, 2025. Knowledge cutoff: Qwen3 235B-A22B Apr 2025.
How do you decide the winner?
This rating uses 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 18, 2026.