Compare / Kimi K2.7 Code vs Kimi K3
Kimi K2.7 Code vs Kimi K3
Which is cheaper, which is more capable, and which does better at each kind of work, from 14 results on 4 sources that measured both models.
Of the 14 results both models have, the better value on each. Results whose reported ranges overlap are too close to call.
Results as of 16 November 2026 (catalogue release 2026-11-16-49bdfa045dee).
Kimi K2.7 Code or Kimi K3? The short answer
Pick Kimi K3 if quality matters most: it scores higher (Intelligence Index 43.6 against 25.8). Pick Kimi K2.7 Code if cost matters more: it costs 3.5 times less per token.
- Kimi K2.7 Code is cheaper: $1.71 against $6.00 per million tokens, 3.5 times less.
- Kimi K3 scores higher on the Artificial Analysis Intelligence Index: 43.6 against 25.8.
- Kimi K3 is newer: released 16 Jul 2026, against 12 Jun 2026 for Kimi K2.7 Code.
- On third-party benchmarks, on the results both models have: Kimi K3 does better for agents and tool use, professional work, reasoning and knowledge and sticking to the facts.
Head to head
Every fact side by side
| Kimi K2.7 Code | Kimi K3 | |
|---|---|---|
| Price per million tokens | $0.95 in · $4.00 out | $3.00 in · $15.00 out |
| Blended (3 in : 1 out) | $1.71 | $6.00 |
| Cheapest host, blended | $1.20 (CoreWeave, INT4) | $2.47 (Relace, FP4) |
| Cached input, per million | $0.19 cached input | $0.30 cached input |
| Intelligence Index | 25.8 | 43.6 |
| Spring Prompt overall | – | 47 |
| Output speed, tokens a second | 54 | 41 |
| Context window | 262,144 tokens | 1,048,576 tokens |
| Released | 12 Jun 2026 | 16 Jul 2026 (newer) |
| Weights | Open | Open |
| Developer based in | China | China |
Prices are the developer's list price where we have read it, otherwise the typical host on OpenRouter; see every model's price. Intelligence and speed from Artificial Analysis.
Which is better for what
Each model's rank among the models on sale today, for each kind of work, from our own benchmarks and licensed sources. Further right is better; the better rank is in green.
From our research
Every shared result
Each model's best result on every measure both have, grouped by benchmark, with the setting that produced it. Under each measure, every model that benchmark has measured is a grey tick, best to the right, so you can see where the two sit in the field. The headline measures of each benchmark come first; the rest are under "Show all". Green marks the better value; where a source reports ranges that overlap, neither is marked.
| Measure | Kimi K2.7 Code | Kimi K3 |
|---|---|---|
| APEX-Agents reported by APEX-Agents · 1 result | ||
| Tasks passed% of tasks, higher is better 31 · 21 of 37 measured | 37.6% | 50.6% |
| Artificial Analysis reported by Artificial Analysis · 3 results | ||
| Artificial Analysis Intelligence Indexindex score, higher is better 75 · 19 of 249 measured | 25.8 | 43.6max reasoning |
| GPQA Diamond% of questions, higher is better 47 · 10 of 232 measured | 89.6% | 93.5%max reasoning |
| Humanity's Last Exam% of questions, higher is better 65 · 20 of 250 measured | 35.0% | 46.9%max reasoning |
| SimpleQA Verified (Epoch AI) reported by SimpleQA Verified (Epoch AI) · 1 result | ||
| Correct answers% of questions, higher is better 47 · 23 of 76 measured | 36.5% | 50.6%max reasoning |
| Vending-Bench 2 reported by Vending-Bench 2 · 1 result | ||
| Money after a yearUS dollars, higher is better 27 · 24 of 58 measured | $5,083 | $5,165 |
Show all 14 results (8 more)
| Measure | Kimi K2.7 Code | Kimi K3 |
|---|---|---|
| Artificial Analysis reported by Artificial Analysis · 8 results | ||
| Terminal-Bench 2.1% of tasks, higher is better 38 · 12 of 131 measured | 67.4% | 85.0%max reasoning |
| Terminal-Bench 4.0% of tasks, higher is better 59 · 37 of 120 measured | 1.0% | 12.6%low reasoning |
| τ-bench banking% of tasks, higher is better 50 · 10 of 121 measured | 20.2% | 46.0%max reasoning |
| Artificial Analysis Coding Indexindex score, higher is better 38 · 10 of 130 measured | 60.8 | 76.2max reasoning |
| Long-context reasoning (AA-LCR)% of questions, higher is better 53 · 1 of 241 measured | 79.3% | 88.7%max reasoning |
| Output speedtokens per second, higher is better 66 · 74 of 78 measured | 53.7 | 41.0max reasoning |
| SciCode% of problems, higher is better 63 · 11 of 121 measured | 47.8% | 59.5%max reasoning |
| Time to first answer tokenseconds, lower is better 71 · 76 of 78 measured | 42.7 | 51.8max reasoning |
Quick answers
Is Kimi K2.7 Code better than Kimi K3?
Pick Kimi K3 if quality matters most: it scores higher (Intelligence Index 43.6 against 25.8). Pick Kimi K2.7 Code if cost matters more: it costs 3.5 times less per token.
Which is cheaper, Kimi K2.7 Code or Kimi K3?
Kimi K2.7 Code is cheaper: $1.71 against $6.00 per million tokens, 3.5 times less.
Which is faster, Kimi K2.7 Code or Kimi K3?
Kimi K2.7 Code writes about 54 tokens a second on its usual API, against 41 for Kimi K3 (Artificial Analysis).
Which is better for agents and tool use, Kimi K2.7 Code or Kimi K3?
Kimi K3 ranks 40 of 159 models on sale for multi-step tasks with tools: support desks, coding agents, function calls, against 69 for Kimi K2.7 Code.
Which is better for professional work, Kimi K2.7 Code or Kimi K3?
Kimi K3 ranks 24 of 51 models on sale for real tasks from banking, consulting and law, business documents and freelance projects, against 41 for Kimi K2.7 Code.
Which is better for reasoning and knowledge, Kimi K2.7 Code or Kimi K3?
Kimi K3 ranks 18 of 187 models on sale for hard questions across science, maths and general knowledge, against 65 for Kimi K2.7 Code.
Which is better for sticking to the facts, Kimi K2.7 Code or Kimi K3?
Kimi K3 ranks 27 of 139 models on sale for summarising without inventing things, and factual answers, against 90 for Kimi K2.7 Code.
Which has the bigger context window, Kimi K2.7 Code or Kimi K3?
Kimi K3: 1,048,576 tokens, against 262,144 for Kimi K2.7 Code.
Run Kimi K2.7 Code and Kimi K3 on your own prompt.
Benchmarks aren't your task. Test both side by side on your prompt, with every answer checked against your rules and the cost of each one.