Qwen3.8 27B vs MiniMax-M3 vs Inkling Small

It is close. Our weighted score puts them within a point (Inkling Small 70/100, MiniMax-M3 69/100, Qwen3.8 27B 67/100), so choose by what matters most for your work: Inkling Small for raw capability and MiniMax-M3 on price. The score weighs capability 50%, price 25%, inputs & features 15%, context window 10%.

MiniMax-M3 is cheaper at $0.30 input / $1.20 output per million tokens (official MiniMax (minimax.io) API price). Inkling Small costs $0.50 input / $1.20 output per million tokens (median across 11 API providers); Qwen3.8 27B costs $0.40 input / $2.50 output per million tokens (median across 39 API providers). At a typical mix of three input tokens to one output token, that is $0.525 per million tokens for MiniMax-M3 versus $0.675 for Inkling Small (1.3× as much) and $0.925 for Qwen3.8 27B (1.8× as much).

Inkling Small scores higher on the Capabilities Index (ECI): Inkling Small 150.2 (#47 of 148), Qwen3.8 27B 149.4 (#53 of 148) and MiniMax-M3 147.0 (#62 of 148). The confidence ranges of the top two overlap (147.5–152.1 vs 147.5–151.6), so treat the gap as small.

There are no published SWE-bench Verified results for Qwen3.8 27B, MiniMax-M3 and Inkling Small yet, so there is no like-for-like coding score. On overall capability, Inkling Small leads, which tends to carry over to coding, but test on your own codebase. All three support tool calling for agent workflows.

MiniMax-M3 and Inkling Small have the largest context windows (1,048,576 and 1,048,576 tokens), against 262,144 for Qwen3.8 27B. Maximum output per response: Qwen3.8 27B up to 32,768, MiniMax-M3 up to 512,000, Inkling Small up to 1,048,576 tokens.

Qwen3.8 27B accepts text, images and video; MiniMax-M3 accepts text, images and video; Inkling Small accepts text, images and audio. They handle the same number of input types.

Yes, all three publish their weights (Apache-2.0), so you can self-host them.

Qwen3.8 27B is the newest, released Aug 14, 2026. Inkling Small came out Jul 30, 2026; MiniMax-M3 came out Jun 1, 2026.

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.