DeepSeek-R1 vs Kimi K2 Instruct vs Qwen3 235B-A22B Thinking 2507
Qwen3 235B-A22B Thinking 2507 comes out ahead, 60 to 56 and 55 on our weighted score, and it is the cheaper option too.
Price in rating: Excluded · Include price
DeepSeek
DeepSeek-R1
56/100- ECI138.6
- Price$1.02 / $5.14
- Context128K
Moonshot AI
Kimi K2 Instruct
55/100- ECI141.3
- Price$0.57 / $2.30
- Context131K
- Our pick
Alibaba (Qwen)
Qwen3 235B-A22B Thinking 2507
60/100- ECI143.3
- Price$0.30 / $2.90
- Context262K
Qwen3 235B-A22B Thinking 2507 is our pick
Qwen3 235B-A22B Thinking 2507 is the better all-round choice, scoring 60/100 against DeepSeek-R1 (56) and Kimi K2 Instruct (55). It leads on capability and context window. The score weighs capability 67%, inputs & features 20%, context window 13%.
- CapabilityQwen3 235B-A22B Thinking 2507Capabilities Index (ECI): Qwen3 235B-A22B Thinking 2507 143.3 · Kimi K2 Instruct 141.3 · DeepSeek-R1 138.6
- Lowest priceQwen3 235B-A22B Thinking 2507Qwen3 235B-A22B Thinking 2507 $0.95 · Kimi K2 Instruct $1.00 · DeepSeek-R1 $2.05 per 1M tokens (3:1 blend)
- Longest contextQwen3 235B-A22B Thinking 2507Qwen3 235B-A22B Thinking 2507 262,144 · Kimi K2 Instruct 131,072 · DeepSeek-R1 128,000 tokens
- Widest inputsSame inputsDeepSeek-R1: Text · Kimi K2 Instruct: Text · Qwen3 235B-A22B Thinking 2507: Text
- Self-hostingAll open weightsEvery model here can be downloaded and run on your own hardware
| Measure | Weight | DeepSeek-R1 | Kimi K2 Instruct | Qwen3 235B-A22B Thinking 2507 |
|---|---|---|---|---|
| CapabilityCapabilities Index (ECI) | 67% | 67 | 69 | 71 |
| Inputs & features | 20% | 35 | 25 | 35 |
| Context window | 13% | 32 | 32 | 43 |
| Overall | 100% | 56/100 | 55/100 | 60/100 |
Every spec in one table
Highlighted cells lead their row. Dashes mean the data is not published.
| Specification | |||
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | 138.6 | 141.3 | 143.3 (best) |
| ECI rank | #118 of 158 | #108 of 158 | #100 of 158 (best) |
| GPQA DiamondGraduate-level science questions | 71.7% | — | 80.1% (best) |
| OTIS Mock AIME 2024–2025Competition mathematics | 53.3% | — | 86.7% (best) |
| SimpleQA VerifiedShort factual questions | — | — | 40.4% |
| Price per million tokens | |||
| Input | $1.02 | $0.57 | $0.30 (best) |
| Output | $5.14 | $2.30 (best) | $2.90 |
| Cached input | — | — | — |
| Blended (3:1) | $2.05 | $1.00 | $0.95 (best) |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Median of 12 providers | Median of 11 providers | Median of 11 providers |
| Limits | |||
| Context window | 128,000 tokens | 131,072 tokens | 262,144 tokens (best) |
| Max output | 32,768 tokens | 32,768 tokens | 81,920 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 | OpenApache 2.0 |
| API model ID | — | — | — |
| API providers | 13 (best) | 11 | 12 |
| Released | Jan 20, 2025 | Jul 11, 2025 | Jul 25, 2025 |
| Knowledge cutoff | Jul 2024 | — | — |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
DeepSeek-R1$20.53
Kimi K2 Instruct$10.30
Qwen3 235B-A22B Thinking 2507$8.80
Which should you choose?
Which is better: DeepSeek-R1, Kimi K2 Instruct or Qwen3 235B-A22B Thinking 2507?
Qwen3 235B-A22B Thinking 2507 is the better all-round choice, scoring 60/100 against DeepSeek-R1 (56) and Kimi K2 Instruct (55). It leads on capability and context window. The score weighs capability 67%, inputs & features 20%, context window 13%.
Which is cheaper, DeepSeek-R1, Kimi K2 Instruct or Qwen3 235B-A22B Thinking 2507?
Qwen3 235B-A22B Thinking 2507 is cheaper at $0.30 input / $2.90 output per million tokens (median across 11 API providers). Kimi K2 Instruct costs $0.57 input / $2.30 output per million tokens (median across 11 API providers); DeepSeek-R1 costs $1.02 input / $5.14 output per million tokens (median across 12 API providers). At a typical mix of three input tokens to one output token, that is $0.95 per million tokens for Qwen3 235B-A22B Thinking 2507 versus $1.00 for Kimi K2 Instruct (1.1× as much) and $2.05 for DeepSeek-R1 (2.2× as much).
Which scores higher on benchmarks?
Qwen3 235B-A22B Thinking 2507 scores higher on the Capabilities Index (ECI): Qwen3 235B-A22B Thinking 2507 143.3 (#100 of 158), Kimi K2 Instruct 141.3 (#108 of 158) and DeepSeek-R1 138.6 (#118 of 158). The confidence ranges of the top two overlap (141.4–145.1 vs 138.3–144.0), so treat the gap as small.
Which is better for coding?
There are no published SWE-bench Verified results for DeepSeek-R1, Kimi K2 Instruct and Qwen3 235B-A22B Thinking 2507 yet, so there is no like-for-like coding score. On overall capability, Qwen3 235B-A22B Thinking 2507 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 235B-A22B Thinking 2507 has the largest context window at 262,144 tokens, against 131,072 for Kimi K2 Instruct and 128,000 for DeepSeek-R1. Maximum output per response: DeepSeek-R1 up to 32,768, Kimi K2 Instruct up to 32,768, Qwen3 235B-A22B Thinking 2507 up to 81,920 tokens.
Which can read images, PDFs, audio or video?
DeepSeek-R1 accepts text; Kimi K2 Instruct accepts text; Qwen3 235B-A22B Thinking 2507 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 Apache 2.0), so you can self-host them.
Which is newer?
Qwen3 235B-A22B Thinking 2507 is the newest, released Jul 25, 2025. Kimi K2 Instruct came out Jul 11, 2025; DeepSeek-R1 came out Jan 20, 2025. Knowledge cutoff: DeepSeek-R1 Jul 2024.
How do you decide the winner?
This rating uses capability (67%, independent benchmark results); inputs & features (20%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (13%, 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.