GLM-5 vs Kimi K2 Thinking vs Qwen3.6 27B

Qwen3.6 27B is the better all-round choice, scoring 65/100 against Kimi K2 Thinking (58) and GLM-5 (55). It leads on inputs & features. Kimi K2 Thinking wins on price. The score weighs capability 50%, price 25%, inputs & features 15%, context window 10%.

Kimi K2 Thinking is cheaper at $0.60 input / $2.50 output per million tokens (median across 10 API providers). Qwen3.6 27B costs $0.60 input / $3.60 output per million tokens (official Alibaba API price); GLM-5 costs $1.00 input / $3.20 output per million tokens (official Z.AI API price). At a typical mix of three input tokens to one output token, that is $1.07 per million tokens for Kimi K2 Thinking versus $1.35 for Qwen3.6 27B (1.3× as much) and $1.55 for GLM-5 (1.4× as much).

Qwen3.6 27B scores higher on the Capabilities Index (ECI): Qwen3.6 27B 146.5 (#68 of 148), Kimi K2 Thinking 146.0 (#72 of 148) and GLM-5 145.8 (#74 of 148). The confidence ranges of the top two overlap (144.2–147.9 vs 143.4–147.6), so treat the gap as small. On individual benchmarks: GPQA Diamond — GLM-5 87.8%, Qwen3.6 27B 85.9%, Kimi K2 Thinking 84.2%; OTIS Mock AIME 2024–2025 — Qwen3.6 27B 91.1%, Kimi K2 Thinking 83.1%, GLM-5 80.0%.

There are no published SWE-bench Verified results for Kimi K2 Thinking and Qwen3.6 27B yet, so there is no like-for-like coding score. On overall capability, Qwen3.6 27B leads, which tends to carry over to coding, but test on your own codebase. All three support tool calling for agent workflows.

Kimi K2 Thinking and Qwen3.6 27B have the largest context windows (262,144 and 262,144 tokens), against 204,800 for GLM-5. Maximum output per response: GLM-5 up to 131,072, Kimi K2 Thinking up to 262,144, Qwen3.6 27B up to 65,536 tokens.

GLM-5 accepts text; Kimi K2 Thinking accepts text; Qwen3.6 27B accepts text, images, audio and video. Qwen3.6 27B handles the widest range of inputs.

Yes, all three publish their weights, so you can self-host them.

Qwen3.6 27B is the newest, released Apr 22, 2026. GLM-5 came out Feb 12, 2026; Kimi K2 Thinking came out Nov 6, 2025. Knowledge cutoff: Kimi K2 Thinking Aug 2024.

Each model gets a 0–100 score on capability (50%, independent benchmark results); price (25%, blended price per million tokens (3 input : 1 output), log scale); inputs & features (15%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (10%, 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 Oct 4, 2026.