Claude Haiku 4.5 vs GLM-5.1
Benchmarks, API pricing and specs, head to head. Data updated 2026-06-10.
Anthropic · Oct 2025
Anthropic's fast tier: Sonnet-4-class coding performance at $1/$5 and more than twice the speed.
Z.ai (Zhipu) · Apr 2026
The April 2026 GLM refresh — the first open-weights model to take #1 on SWE-bench Pro, edging out Claude Opus 4.6.
The verdict
These two models don't yet share verified results on the benchmarks we track, so judge them on specs, pricing and intended use. Claude Haiku 4.5 is about 1.1x cheaper per blended million tokens (3:1 input:output mix). And GLM-5.1 is open-weights (MIT), so it can be self-hosted — a structural advantage if data control or vendor independence matters.
Specs & pricing
| Claude Haiku 4.5 | GLM-5.1 | |
|---|---|---|
| modhub Index | — | — |
| Input price / 1M | $1 | $1.4 |
| Output price / 1M | $5 | $4.4 |
| Context window | 200K | 200K |
| Max output | 64K | 128K |
| Open weights | no | yes (MIT) |
| Reasoning model | yes | yes |
| Multimodal input | text, image | text |
| Knowledge cutoff | Feb 2025 | Feb 2026 |
| Released | Oct 2025 | Apr 2026 |
| Example monthly cost* | $17.50 | $20.60 |
* 10M input + 1.5M output tokens per month at list prices, no caching. Green = better value on that row.
Frequently asked questions
- Which is better, Claude Haiku 4.5 or GLM-5.1?
- These two models don't yet share verified results on the benchmarks we track, so judge them on specs, pricing and intended use. Claude Haiku 4.5 is about 1.1x cheaper per blended million tokens (3:1 input:output mix). And GLM-5.1 is open-weights (MIT), so it can be self-hosted — a structural advantage if data control or vendor independence matters.
- Which is cheaper, Claude Haiku 4.5 or GLM-5.1?
- Claude Haiku 4.5 costs $1/$5 per million input/output tokens, while GLM-5.1 costs $1.4/$4.4. For a typical workload of 10M input and 1.5M output tokens per month, that's $17.50 versus $20.60.
- Which model is better for coding, Claude Haiku 4.5 or GLM-5.1?
- We don't yet track SWE-bench Verified results for both models; check their individual pages for coding-related scores.