GLM-4.5-Air vs GLM-4.7 vs Claude Haiku 4.5

It is close. Our weighted score puts them within a point (GLM-4.5-Air 48/100, Claude Haiku 4.5 47/100, GLM-4.7 41/100), so choose by what matters most for your work: GLM-4.5-Air on price and GLM-4.7 for long inputs. The score weighs price 50%, inputs & features 30%, context window 20%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.

GLM-4.5-Air is cheaper at $0.20 input / $1.10 output per million tokens (official Z.AI API price). GLM-4.7 costs $0.60 input / $2.20 output per million tokens (official Z.AI API price); Claude Haiku 4.5 costs $1.00 input / $5.00 output per million tokens (official Anthropic API price). At a typical mix of three input tokens to one output token, that is $0.425 per million tokens for GLM-4.5-Air versus $1.00 for GLM-4.7 (2.4× as much) and $2.00 for Claude Haiku 4.5 (4.7× as much).

There is no independent benchmark that covers all three models yet. GLM-4.5-Air has not been scored yet, GLM-4.7 has an ECI of 143.7 and Claude Haiku 4.5 has an ECI of 142.2.

There are no published SWE-bench Verified results for GLM-4.5-Air, GLM-4.7 and Claude Haiku 4.5 yet, so there is no like-for-like coding score. Test both on a sample of your own tasks. All three support tool calling for agent workflows.

GLM-4.7 has the largest context window at 204,800 tokens, against 200,000 for Claude Haiku 4.5 and 131,072 for GLM-4.5-Air. Maximum output per response: GLM-4.5-Air up to 98,304, GLM-4.7 up to 131,072, Claude Haiku 4.5 up to 64,000 tokens.

GLM-4.5-Air accepts text; GLM-4.7 accepts text; Claude Haiku 4.5 accepts text, images and PDFs. Claude Haiku 4.5 handles the widest range of inputs.

GLM-4.5-Air and GLM-4.7 publishes its weights and can be self-hosted; Claude Haiku 4.5 is proprietary.

GLM-4.7 is the newest, released Dec 22, 2025. Claude Haiku 4.5 came out Oct 15, 2025; GLM-4.5-Air came out Jul 28, 2025. Knowledge cutoff: GLM-4.5-Air Apr 2025, GLM-4.7 Apr 2025, Claude Haiku 4.5 Feb 28, 2025.

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 Nov 2, 2026.