fix: cloudflare-workers-ai: glm-5.3-flash limit 1310720 exceeds gateway max 1048576 (same as #4422 for glm-5.2)

[github-actions]

Summary

Cloudflare Workers AI @cf/zai-org/glm-5.3-flash had limit.context / limit.output set to 1,310,720, which comes from catalog/search data and is above AI Gateway’s max (max_completion_tokens ≤ 1,048,576). Clients that derive max_tokens from catalog output (e.g. pi) then get HTTP 400. Limits are corrected to match Cloudflare’s documented context window.

Changes

File: providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3-flash.toml

Field	Old	New
limit.context	1_310_720	1_048_576
limit.output	1_310_720	1_048_576

Leading comment notes the gateway cap and docs source.

Evidence

• Cloudflare Workers AI – glm-5.3-flash: Context Window = 1,048,576 tokens; pricing matches existing TOML ($0.15 / $0.50 / $0.03 cache).
• Issue cloudflare-workers-ai: glm-5.3-flash limit 1310720 exceeds gateway max 1048576 (same as #4422 for glm-5.2) #5824 AIG error: max_completion_tokens is too large: 1298829. This model supports at most 1048576 — confirms gateway output max is 1,048,576, not 1,310,720.
• Same class of fix as fix(cloudflare-workers-ai): correct GLM 5.2 token limits #4422 for glm-5.2 (catalog inflated limit → host-safe value).

Lab models/zhipuai/glm-5.3-flash.toml (context = 1_000_000, output = 131_072) was left unchanged; this is a Cloudflare host constraint, not a lab metadata correction.

Validation

• Compared provider TOML to Cloudflare first-party model page (context 1,048,576).
• Cross-checked issue AIG error for the 1,048,576 completion-token ceiling.
• Did not run bun validate or live API calls (no shell per instructions).

Review notes

• Did not change providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.3.toml, which still has 1_310_720 / 1_310_720 and the same Cloudflare docs pattern (1,048,576 context). Out of scope for cloudflare-workers-ai: glm-5.3-flash limit 1310720 exceeds gateway max 1048576 (same as #4422 for glm-5.2) #5824; may need a follow-up.
• Output set equal to context (1,048,576) to match the gateway max reported in the error; Cloudflare docs only publish context window, not a separate max output.

Closes #5824

Automated by the issue fixer: https://github.com/anomalyco/models.dev/actions/runs/33266781527

[jpenilla]

it seems like this got partially reverted by the auto update - pi should be ok now but just flagging for posterity