DeepSeek V3 0324 vs Kimi K2 Thinking vs QwQ 32B
Kimi K2 Thinking comes out ahead, 61 to 54 and 52 on our weighted score, though DeepSeek V3 0324 is 2.2× cheaper per token.
Price in rating: Excluded · Include price
DeepSeek
DeepSeek V3 0324
52/100- ECI135.3
- Price$0.275 / $1.12
- Context164K
- Our pick
Moonshot AI
Kimi K2 Thinking
61/100- ECI145.5
- Price$0.60 / $2.50
- Context262K
Alibaba (Qwen)
QwQ 32B
54/100- ECI136.5
- Price$0.53 / $0.70
- Context131K
Kimi K2 Thinking is our pick
Kimi K2 Thinking is the better all-round choice, scoring 61/100 against QwQ 32B (54) and DeepSeek V3 0324 (52). It leads on capability and context window. The score weighs capability 67%, inputs & features 20%, context window 13%.
- CapabilityKimi K2 ThinkingCapabilities Index (ECI): Kimi K2 Thinking 145.5 · QwQ 32B 136.5 · DeepSeek V3 0324 135.3
- Lowest priceDeepSeek V3 0324DeepSeek V3 0324 $0.486 · QwQ 32B $0.573 · Kimi K2 Thinking $1.07 per 1M tokens (3:1 blend)
- Longest contextKimi K2 ThinkingKimi K2 Thinking 262,144 · DeepSeek V3 0324 163,840 · QwQ 32B 131,072 tokens
- Widest inputsSame inputsDeepSeek V3 0324: Text · Kimi K2 Thinking: Text · QwQ 32B: Text
- Self-hostingAll open weightsEvery model here can be downloaded and run on your own hardware
| Measure | Weight | DeepSeek V3 0324 | Kimi K2 Thinking | QwQ 32B |
|---|---|---|---|---|
| CapabilityCapabilities Index (ECI) | 67% | 63 | 73 | 65 |
| Inputs & features | 20% | 25 | 35 | 35 |
| Context window | 13% | 36 | 43 | 32 |
| Overall | 100% | 52/100 | 61/100 | 54/100 |
Every spec in one table
Highlighted cells lead their row. Dashes mean the data is not published.
| Specification | |||
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | 135.3 | 145.5 (best) | 136.5 |
| ECI rank | #125 of 158 | #90 of 158 (best) | #122 of 158 |
| GPQA DiamondGraduate-level science questions | 67.6% | 84.2% (best) | 65.3% |
| OTIS Mock AIME 2024–2025Competition mathematics | 37.8% | 83.1% (best) | 59.2% |
| Price per million tokens | |||
| Input | $0.275 (best) | $0.60 | $0.53 |
| Output | $1.12 | $2.50 | $0.70 (best) |
| Cached input | — | — | — |
| Blended (3:1) | $0.486 (best) | $1.07 | $0.573 |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Median of 13 providers | Median of 14 providers | Median of 2 providers |
| Limits | |||
| Context window | 163,840 tokens | 262,144 tokens (best) | 131,072 tokens |
| Max output | 163,840 tokens | 262,144 tokens (best) | 8,192 tokens |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | No | No | No |
| PDFs | No | No | No |
| Audio | No | No | No |
| Video | No | No | No |
| Reasoning | No | Yes | Yes |
| Tool calling | Yes | Yes | Yes |
| Structured output | No | No | No |
| Availability | |||
| Weights | Open | Open | Open |
| API model ID | — | — | — |
| API providers | 13 | 15 (best) | 2 |
| Released | Mar 24, 2025 | Nov 6, 2025 | Mar 5, 2025 |
| Knowledge cutoff | — | Aug 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.
DeepSeek V3 0324$4.99
Kimi K2 Thinking$11.00
QwQ 32B$6.70
Which should you choose?
Which is better: DeepSeek V3 0324, Kimi K2 Thinking or QwQ 32B?
Kimi K2 Thinking is the better all-round choice, scoring 61/100 against QwQ 32B (54) and DeepSeek V3 0324 (52). It leads on capability and context window. The score weighs capability 67%, inputs & features 20%, context window 13%.
Which is cheaper, DeepSeek V3 0324, Kimi K2 Thinking or QwQ 32B?
DeepSeek V3 0324 is cheaper at $0.275 input / $1.12 output per million tokens (median across 13 API providers). QwQ 32B costs $0.53 input / $0.70 output per million tokens (median across 2 API providers); Kimi K2 Thinking costs $0.60 input / $2.50 output per million tokens (median across 14 API providers). At a typical mix of three input tokens to one output token, that is $0.486 per million tokens for DeepSeek V3 0324 versus $0.573 for QwQ 32B (1.2× as much) and $1.07 for Kimi K2 Thinking (2.2× as much).
Which scores higher on benchmarks?
Kimi K2 Thinking scores higher on the Capabilities Index (ECI): Kimi K2 Thinking 145.5 (#90 of 158), QwQ 32B 136.5 (#122 of 158) and DeepSeek V3 0324 135.3 (#125 of 158). Their confidence ranges do not overlap (143.7–147.3 vs 133.7–139.2), so the gap is a real one. On individual benchmarks: GPQA Diamond — Kimi K2 Thinking 84.2%, DeepSeek V3 0324 67.6%, QwQ 32B 65.3%; OTIS Mock AIME 2024–2025 — Kimi K2 Thinking 83.1%, QwQ 32B 59.2%, DeepSeek V3 0324 37.8%.
Which is better for coding?
There are no published SWE-bench Verified results for DeepSeek V3 0324, Kimi K2 Thinking and QwQ 32B yet, so there is no like-for-like coding score. On overall capability, Kimi K2 Thinking 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?
Kimi K2 Thinking has the largest context window at 262,144 tokens, against 163,840 for DeepSeek V3 0324 and 131,072 for QwQ 32B. Maximum output per response: DeepSeek V3 0324 up to 163,840, Kimi K2 Thinking up to 262,144, QwQ 32B up to 8,192 tokens.
Which can read images, PDFs, audio or video?
DeepSeek V3 0324 accepts text; Kimi K2 Thinking accepts text; QwQ 32B accepts text. They handle the same number of input types.
Are any of these open source?
Yes, all three publish their weights, so you can self-host them.
Which is newer?
Kimi K2 Thinking is the newest, released Nov 6, 2025. DeepSeek V3 0324 came out Mar 24, 2025; QwQ 32B came out Mar 5, 2025. Knowledge cutoff: Kimi K2 Thinking Aug 2024, QwQ 32B Apr 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.