LLMs TigerLive LLM benchmarks

Kimi K2.7 Code: benchmark scores, price and ranking

Kimi K2.7 Code from Moonshot ranks #444 of 743 LLMs on the LLMs Tiger aggregate benchmark score (47.3), based on 26 benchmark results captured up to 2026-10-13.

Organisation
Moonshot
Type
Open-weight / local model
Open weights
Yes
Released
2026-06-12 (source: epoch)
Cheapest API price
$0.671 in / $3.35 out per million tokens (via openrouter, observed 2026-10-13T19:47Z)
Aggregate rank
#444 of 743 (score 47.3)

Relative to other models, Kimi K2.7 Code is strongest on Epoch ECI (#9 of 204) and weakest on LiveBench Data Analysis (#17 of 17).

Across the 25 benchmarks where at least ten models are scored, Kimi K2.7 Code is in the top 10% on 1 and in the top half on 13.

Its cheapest listed input price, $0.671 per million tokens, is lower than that of 42% of the 427 models with a listed price.

Kimi K2.7 Code benchmark results

Kimi K2.7 Code benchmark scores, ranked against all models measured on each benchmark
BenchmarkScoreRankSourceCaptured
Epoch ECI150#9 of 204Epoch AI Benchmarking Hub2026-10-13
OTIS Mock AIME95.6%#26 of 189Epoch AI Benchmarking Hub2026-10-13
SimpleBench57.9%#11 of 61Epoch AI Benchmarking Hub2026-10-13
GPQA Diamond87.9%#51 of 275Epoch AI Benchmarking Hub2026-10-13
WeirdML54.1%#23 of 109Epoch AI Benchmarking Hub2026-10-13
Vending-Bench 2$5,083#5 of 19Epoch AI Benchmarking Hub2026-10-13
CritPt10.0%#39 of 115Epoch AI Benchmarking Hub2026-10-13
Chess Puzzles21.0%#42 of 122Epoch AI Benchmarking Hub2026-10-13
ALE-Bench886#31 of 79Epoch AI Benchmarking Hub2026-10-13
SciCode47.5%#42 of 106Epoch AI Benchmarking Hub2026-10-13
WebDev Arena1,473#27 of 67Epoch AI Benchmarking Hub2026-10-13
LiveBench Reasoning82.8%#7 of 17LiveBench 2026_06_25 official CSV2026-10-13
Surface Evolver48.8%#6 of 13Epoch AI Benchmarking Hub2026-10-13
FrontierMath T1-354.0%#36 of 68Epoch AI Benchmarking Hub2026-10-13
LiveBench Language77.9%#9 of 17LiveBench 2026_06_25 official CSV2026-10-13
GBA-Eval0.9%#4 of 7Epoch AI Benchmarking Hub2026-10-13
SimpleQA Verified36.5%#40 of 69Epoch AI Benchmarking Hub2026-10-13
LiveBench Agentic Coding45.7%#10 of 17LiveBench 2026_06_25 official CSV2026-10-13
FrontierCode30.1%#9 of 15Epoch AI Benchmarking Hub2026-10-13
LiveBench Coding74.0%#12 of 17LiveBench 2026_06_25 official CSV2026-10-13
APEX-Agents37.6%#13 of 16Epoch AI Benchmarking Hub2026-10-13
LiveBench Mathematics79.6%#14 of 17LiveBench 2026_06_25 official CSV2026-10-13
FrontierMath T412.2%#42 of 50Epoch AI Benchmarking Hub2026-10-13
LiveBench IF56.3%#16 of 17LiveBench 2026_06_25 official CSV2026-10-13
DeepSWE30.5%#19 of 19Epoch AI Benchmarking Hub2026-10-13
LiveBench Data Analysis62.7%#17 of 17LiveBench 2026_06_25 official CSV2026-10-13

About Kimi K2.7 Code

Kimi K2.7 Code is an open-weight model that can run on your own hardware from Moonshot. It was released on 2026-06-12, according to Epoch AI. The cheapest API listing we track charges $0.671 per million input tokens and $3.35 per million output tokens, through openrouter. We hold 26 benchmark results for it across 4 areas, and 1 is in the top tenth of its board.

Where Kimi K2.7 Code stands

On our aggregate score, Kimi K2.7 Code is #444 of 743 ranked models, around the middle of the table. Its score is 47.3; the scale is centred on 50, so it sits 2.7 points below the typical model. That beats 40% of the ranked field. Among the 13 ranked Moonshot models it is #11. The next one up is Kimi K2 Thinking at 47.8. The next one down is Kimi K2 0905 (Novita) at 47.3.

Kimi K2.7 Code, benchmark by benchmark

On Epoch ECI, Kimi K2.7 Code scores 150 and ranks #9 of 204. Only 8 models score higher, led by Qwen 3.8 Max at 156, which is 6 points ahead. It covers Epoch Capabilities Index: Epoch's own aggregate of ~40 benchmarks. The board median is 127, so Kimi K2.7 Code is 23 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

On OTIS Mock AIME, Kimi K2.7 Code reaches 95.6%. It ranks #26 of 189, ahead of 86% of the field, and 4.4 points behind Claude Fable 5.1 (max) at 100.0%. The results either side are GPT-5.2 (xhigh) (96.1%) and Qwen3.7 Max (95.6%). It covers competition maths (AIME 2024-25 style). The board median is 57.8%, so Kimi K2.7 Code is 37.8 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

Kimi K2.7 Code manages 57.9% on SimpleBench, good for #11 of 61. Claude Fable 5 (max) leads at 81.9%, 24.0 points further up, and 82% of models sit below. The results either side are Claude Opus 4 (58.8%) and GPT-5 (high) (56.7%). It covers trick questions that test common sense. The board median is 41.4%, so Kimi K2.7 Code is 16.5 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

GPQA Diamond puts Kimi K2.7 Code at 87.9%, #51 of 275. That beats 81% of the models tested, but 7.9 points short of GPT-6 Astra (max) (95.8%). The results either side are NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4 (87.9%) and Qwen3.7 Plus (87.9%). It covers expert-level science questions. The board median is 71.7%, so Kimi K2.7 Code is 16.2 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

On WeirdML, Kimi K2.7 Code reaches 54.1%. It ranks #23 of 109, ahead of 79% of the field, and 39.2 points behind GPT-6 Astra (max) at 93.3%. The results either side are GPT-5-codex (high) (54.5%) and gemini-2.5-pro_16K (54.0%). It covers unusual machine-learning tasks. The board median is 41.2%, so Kimi K2.7 Code is 12.9 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

Kimi K2.7 Code manages $5,083 on Vending-Bench 2, good for #5 of 19. Kimi K2.6 leads at $6,205, 1,122 points further up, and 74% of models sit below. The results either side are Qwen 3.6 Plus (2026-04-02) ($5,115) and Claude Fable 5 (max) ($4,967). It covers running a vending business for a year (final money, $). The board median is $2,377, so Kimi K2.7 Code is 2,706 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

FrontierMath T4 is a weak spot. Kimi K2.7 Code scores 12.2%, #42 of 50, behind 82% of the models measured. The leader, GPT-6.1 Sol (max), reaches 100.0%. It covers research-level maths, tier 4. The board median is 30.5%, so Kimi K2.7 Code is 18.3 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

LiveBench IF is a weak spot. Kimi K2.7 Code scores 56.3%, #16 of 17, behind 88% of the models measured. The leader, gemini-3.1-pro-preview-high, reaches 79.1%. It covers LiveBench release 2026_06_25. The board median is 65.2%, so Kimi K2.7 Code is 8.9 points below it. Source: LiveBench 2026_06_25 official CSV, captured 2026-10-13.

DeepSWE is a weak spot. Kimi K2.7 Code scores 30.5%, #19 of 19, behind 95% of the models measured. The leader, Gemini 3.8 Flash (high), reaches 73.8%. It covers long-horizon software engineering. The board median is 67.0%, so Kimi K2.7 Code is 36.5 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

LiveBench Data Analysis is a weak spot. Kimi K2.7 Code scores 62.7%, #17 of 17, behind 94% of the models measured. The leader, GPT-5.5 (xhigh), reaches 81.6%. It covers LiveBench release 2026_06_25. The board median is 71.8%, so Kimi K2.7 Code is 9.1 points below it. Source: LiveBench 2026_06_25 official CSV, captured 2026-10-13.

The full table above also lists the other 16 results, including boards with fewer than ten models, where a rank says little.

Strengths and weaknesses

Software engineering: uneven, ahead of 42% of the field on average across 7 benchmarks. Its best board here is ALE-Bench (#31 of 79) and its weakest is DeepSWE (#19 of 19). Reasoning, maths and science: solid, ahead of 57% of the field on average across 8 benchmarks. Its best board here is OTIS Mock AIME (#26 of 189) and its weakest is FrontierMath T4 (#42 of 50). General knowledge and chat: solid, ahead of 51% of the field on average across 9 benchmarks. Its best board here is Epoch ECI (#9 of 204) and its weakest is LiveBench Data Analysis (#17 of 17).

Price and value

The median input price across the 427 priced models is $0.500 per million tokens, so Kimi K2.7 Code costs more than the typical listing. Models that score within 1.5 points or better and cost less: Qwen2.5-Coder-7B-Instruct ($0.010, 49.5), DeepSeek V4.1 Flash (high) ($0.010, 48.9), Qwen2.5-Coder-3B-Instruct ($0.010, 46.5). Similar scores cost more on Mistral Large 4 (high) ($0.680) and Llama 3.1-70B ($0.720).

Release history

The Moonshot model released before it was Kimi K2.6 on 2026-04-20, which scores 48.4. The next one was Kimi K3 (max) on 2026-07-16, at 55.7.

Reading these numbers

The 26 scores on this page come from 2 sources: Epoch AI Benchmarking Hub (19), LiveBench 2026_06_25 official CSV (7). 7 of them come from an official leaderboard and 19 from datasets or third-party runs.

Models ranked nearby

More from Moonshot

Compare it on the live leaderboard → · How the aggregate score works