GPT-4o (2024-11-20) vs Kimi K2 Thinking
Kimi K2 Thinking leads GPT-4o (2024-11-20) on the Artificial Analysis Intelligence Index, 22.0 to 8.4, and Kimi K2 Thinking costs 4.0x less per output token ($2.50 vs $10.00 per million). Updated October 22, 2026; 11 benchmarks have a result for both.
GPT-4o (2024-11-20) or Kimi K2 Thinking: which to pick
Neither model has a coding score yet, so the two cannot be ranked on coding.
Pick GPT-4o (2024-11-20) for work measured by LiveCodeBench (Vals AI) or WeirdML, where it leads.
- Shared results
- 11
- GPT-4o (2024-11-20) only
- 13
- Kimi K2 Thinking only
- 19
- Like-for-like categories
- 5 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
No call
Neither has a coding score yet.
One side not measured
Intelligence
AA Intelligence Index
Kimi K2 Thinking
Kimi K2 Thinking leads, 22.0 against 8.4.
Same board, both measured
Reasoning
GPQA Diamond
Kimi K2 Thinking
Kimi K2 Thinking leads, 78.5% against 53.8%.
Same board, both measured
Math
AIME
Kimi K2 Thinking
Kimi K2 Thinking leads, 85.4% against 11.9%.
Same board, both measured
Knowledge
MMLU Pro
Kimi K2 Thinking
Kimi K2 Thinking leads, 81.1% against 72.6%.
Same board, both measured
Web design
Design Arena: websites
No call
Only Kimi K2 Thinking has a Design Arena: websites result, 1113.
Directional only
Long documents
Listed context window
Kimi K2 Thinking
Kimi K2 Thinking takes 262K tokens against 128K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Kimi K2 Thinking
GPT-4o (2024-11-20) $0.0075, Kimi K2 Thinking $0.0019 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only Kimi K2 Thinking has the number, $0.61.
Directional only
At a glance
- Coding index
- GPT-4o (2024-11-20) not benchmarked, Kimi K2 Thinking not benchmarked.
- Widest category
- Math, 36.9 points apart on average over 2 shared benchmarks.
- Widest single gap
- AIME: GPT-4o (2024-11-20) 11.9% against Kimi K2 Thinking 85.4%, 73.5 points apart on a 0 to 100 scale.
- Output price
- GPT-4o (2024-11-20) $10.00 against Kimi K2 Thinking $2.50 per million output tokens, 4.0x apart.
- Context window
- GPT-4o (2024-11-20) takes 128K tokens, Kimi K2 Thinking 262K.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | GPT-4o (2024-11-20) | Bars | Kimi K2 Thinking | Ties |
|---|---|---|---|---|
| Coding | 0 | 2 | ||
| Intelligence | 0 | 1 | ||
| Reasoning | 0 | 1 | ||
| Math | 1 | 1 | ||
| Knowledge | 2 | 3 | ||
| All shared boards | 3 | 8 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index02
| Benchmark | GPT-4o (2024-11-20) | Kimi K2 Thinking | Leader |
|---|---|---|---|
| Aider polyglot | 18.2% | Not measured | Only GPT-4o (2024-11-20) measured |
| SWE-bench Verified | Not measured | 63.4% | Only Kimi K2 Thinking measured |
| LiveCodeBench (Vals AI) | 43.4% | 63.1% | Kimi K2 Thinking |
| SWE-bench Verified (Vals AI) | Not measured | 60.2% | Only Kimi K2 Thinking measured |
| LiveCodeBench (Artificial Analysis run) | Not measured | 85.3% | Only Kimi K2 Thinking measured |
| Terminal-Bench Hard (Artificial Analysis run) | Not measured | 31.1% | Only Kimi K2 Thinking measured |
| Terminal-Bench | Not measured | 35.7% | Only Kimi K2 Thinking measured |
| GSO-Bench | 0.0% | Not measured | Only GPT-4o (2024-11-20) measured |
| WeirdML | 25.1% | 42.8% | Kimi K2 Thinking |
| Cybench | 12.5% | Not measured | Only GPT-4o (2024-11-20) measured |
| ALE-Bench | Not measured | 597.5 | Only Kimi K2 Thinking measured |
| Neither model has a result on the other 28 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | GPT-4o (2024-11-20) | Kimi K2 Thinking | Leader |
|---|---|---|---|
| AA Intelligence Index | 8.4 | 22.0 | Kimi K2 Thinking |
| Epoch Capabilities Index | Not measured | 145.8 | Only Kimi K2 Thinking measured |
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index3 boards, none shared
| Benchmark | GPT-4o (2024-11-20) | Kimi K2 Thinking | Leader |
|---|---|---|---|
| τ²-Bench Telecom | Not measured | 93.0% | Only Kimi K2 Thinking measured |
| METR time horizons | Not measured | 59.2% | Only Kimi K2 Thinking measured |
| TheAgentCompany | 8.6% | Not measured | Only GPT-4o (2024-11-20) measured |
| Neither model has a result on the other 38 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | GPT-4o (2024-11-20) | Kimi K2 Thinking | Leader |
|---|---|---|---|
| GPQA Diamond | 53.8% | 78.5% | Kimi K2 Thinking |
| MMMU Pro (Vals AI) | 62.2% | Not measured | Only GPT-4o (2024-11-20) measured |
| GPQA Diamond (Artificial Analysis run) | Not measured | 83.8% | Only Kimi K2 Thinking measured |
| Humanity's Last Exam (Artificial Analysis run) | Not measured | 23.8% | Only Kimi K2 Thinking measured |
| GPQA Diamond (Epoch AI run) | 47.9% | Not measured | Only GPT-4o (2024-11-20) measured |
| VPCT | 40.0% | Not measured | Only GPT-4o (2024-11-20) measured |
| Neither model has a result on the other 16 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index11
| Benchmark | GPT-4o (2024-11-20) | Kimi K2 Thinking | Leader |
|---|---|---|---|
| AIME | 11.9% | 85.4% | Kimi K2 Thinking |
| MATH 500 (Vals AI) | 74.0% | Not measured | Only GPT-4o (2024-11-20) measured |
| MGSM (Vals AI) | 90.4% | 90.2% | GPT-4o (2024-11-20) |
| AA Math Index | Not measured | 94.7 | Only Kimi K2 Thinking measured |
| AIME 2025 (Artificial Analysis run) | Not measured | 94.7% | Only Kimi K2 Thinking measured |
| OTIS Mock AIME | 6.3% | Not measured | Only GPT-4o (2024-11-20) measured |
| MATH level 5 | 49.8% | Not measured | Only GPT-4o (2024-11-20) measured |
| Neither model has a result on the other 6 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index23
| Benchmark | GPT-4o (2024-11-20) | Kimi K2 Thinking | Leader |
|---|---|---|---|
| MMLU Pro | 72.6% | 81.1% | Kimi K2 Thinking |
| LegalBench (Vals AI) | 82.2% | 80.2% | GPT-4o (2024-11-20) |
| CaseLaw (Vals AI) | 59.7% | 65.7% | Kimi K2 Thinking |
| CorpFin (Vals AI) | 45.9% | 60.6% | Kimi K2 Thinking |
| TaxEval (Vals AI) | 74.5% | 71.7% | GPT-4o (2024-11-20) |
| MortgageTax (Vals AI) | 57.4% | Not measured | Only GPT-4o (2024-11-20) measured |
| MedQA (Vals AI) | Not measured | 92.6% | Only Kimi K2 Thinking measured |
| MMLU Pro (Artificial Analysis run) | Not measured | 84.8% | Only Kimi K2 Thinking measured |
| GeoBench | 71.0% | Not measured | Only GPT-4o (2024-11-20) measured |
| Neither model has a result on the other 9 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index1 board, none shared
| Benchmark | GPT-4o (2024-11-20) | Kimi K2 Thinking | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1113 | Only Kimi K2 Thinking measured |
| Neither model has a result on the other 8 boards here | |||
GeneralDirectional onlyPublic boards, shown as published, kept out of the index4 boards, none shared
| Benchmark | GPT-4o (2024-11-20) | Kimi K2 Thinking | Leader |
|---|---|---|---|
| IFBench | Not measured | 68.1% | Only Kimi K2 Thinking measured |
| Long context reasoning (AA-LCR) | Not measured | 72.0% | Only Kimi K2 Thinking measured |
| Creative writing (Lech Mazur) | 81.8% | Not measured | Only GPT-4o (2024-11-20) measured |
| CL-bench | Not measured | 17.6% | Only Kimi K2 Thinking measured |
| Neither model has a result on the other 42 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | GPT-4o (2024-11-20) | Kimi K2 Thinking | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
Kimi K2 Thinking costs less
Repository review
50K input + 3K output tokens
Kimi K2 Thinking costs less
Agent loop
200K input + 10K output tokens
Kimi K2 Thinking costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | OpenAI GPT-4o (2024-11-20) | MoonshotAI Kimi K2 Thinking |
|---|---|---|
| Maker | OpenAI | Moonshot AI |
| Weights | No public weights listed | Open weights |
| Released | November 20, 2024 | November 6, 2025 |
| Context window | 128K | 262K |
| Input $/Mtok | $2.50 | $0.60 |
| Output $/Mtok | $10.00 | $2.50 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 0 | 1 |
| Cost per solved task | Not computable | $0.61 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is GPT-4o (2024-11-20) better than Kimi K2 Thinking for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. Pick GPT-4o (2024-11-20) for work measured by LiveCodeBench (Vals AI) or WeirdML, where it leads.
Which is cheaper, GPT-4o (2024-11-20) or Kimi K2 Thinking?
GPT-4o (2024-11-20) costs $2.50 per million input tokens and $10.00 per million output tokens; Kimi K2 Thinking costs $0.60 and $2.50. Cost per solved task is known for Kimi K2 Thinking only.
Which has the bigger context window, GPT-4o (2024-11-20) or Kimi K2 Thinking?
Kimi K2 Thinking, with 262K tokens against 128K.
How many benchmarks is the GPT-4o (2024-11-20) vs Kimi K2 Thinking comparison based on?
11 benchmarks have a result for both models as of October 22, 2026. GPT-4o (2024-11-20)'s coding index rests on 0 independent benchmarks and Kimi K2 Thinking's on 1. Where only one model has a result, the page says so instead of naming a winner.