GLM-5.2
Open flagship GLM for long-horizon coding agents and million-token context work
Encoding fertility 1.20×
Harnesses
| Stack | Passed | Attempts | Time | $ / MU | $ / pass | $ / hr | Tokens / pass | Burn | $ / Zhipu AI |
|---|---|---|---|---|---|---|---|---|---|
| Qwen | 3/3 | 6 | 32m 28s | $0.000595 | $0.8521 | $4.74 | 140,702 | 1.00× | $0.8521 |
Official jobs
Pass or fail on each sealed job. Attempts include retries. Failed tries stay in the bill.
| Durable work queue | JSON Patch (RFC 6902 subset) | Token bucket and sliding window |
|---|---|---|
| pass · 4 | pass · 1 | pass · 1 |
Tokens by job
Billed tokens on each official job, including retries. Cache reads drop out of this count.
Where the bill goes
Failed attempts stay in the bill. Cache hits are a discount, shown separately.
- Uncached in
- Output
- Thinking
- Cache write
Retry tax
Attempts are harness calls, including failed hidden-verifier retries. Extra tries stay in the bill.
- Qwen6 attempts · 3 extra / 3 jobs
Cache hits
Hits divided by uncached input plus hits. No cache means the harness did not report a discount.
- Qwen96%
Encoding by slice
Native tokens per MU on the same frozen basket. Tool schemas are where fertility often doubles.
| Slice | Characters | Native tokens | Fertility | True in | Source |
|---|---|---|---|---|---|
| English prose | 4,474 | 1,040 | 0.93× | $1.30 | estimate |
| Code | 2,321 | 583 | 1.00× | $1.41 | estimate |
| Structured | 3,067 | 646 | 0.84× | $1.18 | estimate |
| Tool schemas | 2,378 | 612 | 1.03× | $1.44 | estimate |
| CJK | 831 | 723 | 3.48× | $4.87 | estimate |
| Instructions | 1,164 | 260 | 0.89× | $1.25 | estimate |
Endpoints
| Provider | SKU | List in | List out | Cache hit | Cache write |
|---|---|---|---|---|---|
Zhipu AI | glm-5.2 | $1.40 | $4.40 | $0.26 | $0.00 |