Model comparison
Kimi K2 (Jul 2025) vs Llama 3.1-70B
Kimi K2 (Jul 2025) is the stronger model overall, scoring 41.2 to 29.6 on the Noometry Index. Llama 3.1-70B costs 2.5× less per token, which makes it the better buy when Kimi K2 (Jul 2025)'s lead doesn't matter for your workload.
Last verified . 25 shared benchmarks.
MetaSummary
- They share 25 benchmarks with published results for both. Kimi K2 (Jul 2025) scores higher in 9 categories and Llama 3.1-70B in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where Kimi K2 (Jul 2025) leads 42.7 to 13.5.
- The biggest single-benchmark swing is Omni-MATH: 65.4% for Kimi K2 (Jul 2025) and 21% for Llama 3.1-70B.
- Llama 3.1-70B is cheaper at $0.40 / $0.40 per million input/output tokens, against $0.57 / $2.30 for Kimi K2 (Jul 2025).
- Kimi K2 (Jul 2025) accepts more context: 262K tokens versus 128K.
Side by side
| Kimi K2 (Jul 2025) | Llama 3.1-70B | |
|---|---|---|
| Provider | Moonshot AI | Meta |
| Noometry Index | 41.2 | 29.6 |
| Released | 2025-07-12 | 2024-07-23 |
| Weights | Open | Open |
| Context window | 262K | 128K |
| Max output | 262K | 4K |
| Input $ / M tokens | $0.57 | $0.40 |
| Output $ / M tokens | $2.30 | $0.40 |
| Results tracked | 42 | 35 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K2 (Jul 2025) leads
Kimi K2 (Jul 2025): 42.4 (#102), Llama 3.1-70B: 30.3 (#296)
| Benchmark | Kimi K2 (Jul 2025) | Llama 3.1-70B |
|---|---|---|
| WeirdML | 42.8% | 9% |
| LMArena Coding | 1399 | 1260 |
| SWE-bench Verified (bash only) | 63.4% | — |
| Aider Polyglot | 59.1% | — |
| GSO | 4.9% | — |
| BigCodeBench Instruct | — | 46.1% |
| BigCodeBench Complete | — | 54.8% |
| ALE-Bench | 597.5 | — |
Agentic & Tool Use Kimi K2 (Jul 2025) leads
Kimi K2 (Jul 2025): 32.4 (#64), Llama 3.1-70B: 25.1 (#112)
| Benchmark | Kimi K2 (Jul 2025) | Llama 3.1-70B |
|---|---|---|
| Terminal-Bench | 35.7% | — |
| Berkeley Function Calling Leaderboard | 59.1% | — |
| TheAgentCompany | — | 6.9% |
| BALROG | — | 27.9% |
| METR Time Horizons | 59.2% | — |
Reasoning Kimi K2 (Jul 2025) leads
Kimi K2 (Jul 2025): 23.3 (#179), Llama 3.1-70B: 21.6 (#220)
| Benchmark | Kimi K2 (Jul 2025) | Llama 3.1-70B |
|---|---|---|
| LMArena Hard Prompts | 1384 | 1241 |
| Epoch Capabilities Index | 146.01 | 125.92 |
| SimpleBench | 26.3% | — |
| Kagi LLM Benchmark | 64.4% | — |
| DTBench | — | 60% |
| LMCA | — | 14.8% |
| ForecastBench | 60.2 | — |
Math Kimi K2 (Jul 2025) leads
Kimi K2 (Jul 2025): 42.7 (#83), Llama 3.1-70B: 13.5 (#304)
| Benchmark | Kimi K2 (Jul 2025) | Llama 3.1-70B |
|---|---|---|
| Omni-MATH | 65.4% | 21% |
| LMArena Math | 1397 | 1252 |
| OTIS Mock AIME 2024-2025 | — | 3.6% |
| MATH Level 5 | — | 36.7% |
| FrontierMath (Feb 2025 set) | 21.4% | — |
| FrontierMath Tier 4 (v1) | 0% | — |
Knowledge Kimi K2 (Jul 2025) leads
Kimi K2 (Jul 2025): 37.3 (#157), Llama 3.1-70B: 24.2 (#269)
| Benchmark | Kimi K2 (Jul 2025) | Llama 3.1-70B |
|---|---|---|
| MMLU-Pro | 81.9% | 65.3% |
| GPQA (HELM) | 65.3% | 42.6% |
| LMArena Expert | 1365 | 1209 |
| GPQA Diamond | — | 44.2% |
| Confabulations | 20.4% | — |
| Vectara Hallucination Rate | 17.9% | — |
| MMLU | — | 80.1% |
Multilingual Kimi K2 (Jul 2025) leads
Kimi K2 (Jul 2025): 49.6 (#130), Llama 3.1-70B: 38.8 (#225)
| Benchmark | Kimi K2 (Jul 2025) | Llama 3.1-70B |
|---|---|---|
| LMArena Non-English | 1372 | 1219 |
| LMArena Chinese | 1415 | 1215 |
| LMArena French | 1379 | 1261 |
| LMArena German | 1387 | 1222 |
| LMArena Japanese | 1349 | 1132 |
| LMArena Korean | 1325 | 1140 |
| LMArena Russian | 1385 | 1234 |
| LMArena Spanish | 1386 | 1253 |
Instruction Following Kimi K2 (Jul 2025) leads
Kimi K2 (Jul 2025): 71.1 (#156), Llama 3.1-70B: 65.3 (#223)
| Benchmark | Kimi K2 (Jul 2025) | Llama 3.1-70B |
|---|---|---|
| IFEval | 85% | 82.1% |
| LMArena Instruction Following | 1348 | 1231 |
Long Context Kimi K2 (Jul 2025) leads
Kimi K2 (Jul 2025): 41.2 (#145), Llama 3.1-70B: 37.6 (#214)
| Benchmark | Kimi K2 (Jul 2025) | Llama 3.1-70B |
|---|---|---|
| LMArena Longer Query | 1353 | 1241 |
| Fiction.LiveBench | 66.7% | — |
| CL-bench | 17.6% | — |
Writing & Preference Kimi K2 (Jul 2025) leads
Kimi K2 (Jul 2025): 62.3 (#78), Llama 3.1-70B: 35.4 (#267)
| Benchmark | Kimi K2 (Jul 2025) | Llama 3.1-70B |
|---|---|---|
| LMArena Text | 1380 | 1261 |
| LMArena Creative Writing | 1350 | 1232 |
| EQ-Bench Creative Writing | 1666 | 784 |
| WildBench | 86.2% | 75.8% |
| LMArena Multi-Turn | 1371 | 1256 |
| Short-Story Creative Writing | 85.6% | — |
Frequently asked questions
Is Kimi K2 (Jul 2025) better than Llama 3.1-70B?
Kimi K2 (Jul 2025) is the stronger model overall, scoring 41.2 to 29.6 on the Noometry Index. Llama 3.1-70B costs 2.5× less per token, which makes it the better buy when Kimi K2 (Jul 2025)'s lead doesn't matter for your workload.
Which is cheaper, Kimi K2 (Jul 2025) or Llama 3.1-70B?
Llama 3.1-70B is cheaper. It lists at $0.40 per million input tokens and $0.40 per million output tokens; Kimi K2 (Jul 2025) lists at $0.57 and $2.30.
Is Kimi K2 (Jul 2025) or Llama 3.1-70B better for coding?
Kimi K2 (Jul 2025) scores higher on coding benchmarks: 42.4 versus 30.3 in the Noometry coding category.
Which has the bigger context window?
Kimi K2 (Jul 2025) does, with 262K tokens against 128K.
How many benchmarks do Kimi K2 (Jul 2025) and Llama 3.1-70B share?
25 benchmarks have published results for both models. Kimi K2 (Jul 2025) has 42 scored results on Noometry and Llama 3.1-70B has 35.