Z.ai's GLM-5.3 is positioned as a coding-value model and costs less than Claude Sonnet 5.5 per input and output token. Sonnet's cache-hit price is lower, which matters for the agent workloads GLM targets.
GLM-5.3 is about 30% cheaper on input and 2.3× cheaper on output than Claude Sonnet 5.5. Sonnet 5.5's cache hit ($0.10) is well under GLM's ($0.26), so on cache-heavy coding agents the savings shrink to around a fifth.
| GLM-5.3 | Claude Sonnet 5.5 | |
|---|---|---|
| Provider | Z.ai | Anthropic |
| Tier | Mid-tier | Mid-tier |
| Input / 1M tokens | $1.40 | $2.00 |
| Output / 1M tokens | $4.40 | $10.00 |
| Cache hit / 1M | $0.26 | $0.10 |
| Cache write / 1M | — | $2.50 |
| Context window | 1M | 1M |
| Notes | GLM flagship; replaced 5.2 at same price; strong coding value | Released Sep 28, 2026, replacing Sonnet 5 at the identical $2/$10 per 1M price; Sonnet 5 is now a legacy model |
GLM-5.3 is 30% cheaper on input tokens ($1.40 vs $2.00 per 1M). GLM-5.3 is 2.3× cheaper on output tokens ($4.40 vs $10.00 per 1M).
| Workload | GLM-5.3 | Claude Sonnet 5.5 | Difference |
|---|---|---|---|
| Support chatbot 10,000 conversations × (2,000 input + 400 output tokens) | $45.60 | $80.00 | GLM-5.3 saves $34.40 |
| Document Q&A 1,000 questions × (30,000 input + 800 output tokens) | $45.52 | $68.00 | GLM-5.3 saves $22.48 |
| Coding agent session 4M input tokens (90% cache hits) + 150,000 output tokens | $2.16 | $2.66 | GLM-5.3 saves $0.50 |
Standard rates per 1M tokens from data on this site. The coding-agent row assumes 90% of input is read from the prompt cache (full input price where a model has no cache rate). Cache-write fees, batch discounts and long-context surcharges listed in the notes are not included. Updated November 2, 2026.
you want lower output costs on coding and generation work, and your agent's context isn't mostly cached.
your agents re-read a large cached context, you need Anthropic's Batch API or data-residency options, or your evals favour it.
Coding agents re-send the same files and tool definitions each turn, so cached input is often most of the token count. Sonnet 5.5's cache hit is 5% of its input price; GLM-5.3's is about a fifth of its input. Z.ai currently lists cached-input storage as limited-time free, while Anthropic charges $2.50 per 1M tokens to write a 5-minute cache. Run your own trace through both rate cards: the more output-heavy the session, the more GLM saves.
Anthropic bills Sonnet 5.5's full 1M window at standard rates. This site lists GLM-5.3 at 1M with no long-context tier; Z.ai's pricing page doesn't state its context length, so confirm limits in its model docs before sending very long prompts.
Anthropic's Batch API halves Sonnet 5.5's input and output rates and stacks with caching. Batched Sonnet output is close to GLM-5.3's standard output rate, which makes Sonnet competitive for overnight jobs.
Yes on standard rates: $1.40 vs $2.00 input and $4.40 vs $10.00 output per 1M tokens. Sonnet 5.5 has the cheaper cache hit.
GLM-5.3 replaced GLM-5.2 as Z.ai's GLM flagship, at the same price.
Z.ai's pricing page lists cached-input storage for GLM-5.3 as limited-time free.
How we choose: prices come from each vendor's official pricing page and are re-checked every week. We take no affiliate fees, so rankings aren't for sale.