fix(models): refresh model pricing

[patlux]

Summary

• refresh display pricing for all 51 models in the current Command Code provider catalog
• remove expired Qwen promotional pricing and add current free, discounted, and newly available models
• support request-wide context pricing tiers and one-hour cache-write pricing using pi's current cost semantics
• replace source-regex assertions with imported pricing data and dated catalog/pricing snapshots
• fail tests when temporary GPT-5.6 Terra/Luna or Claude Sonnet 5 pricing needs review

Sources

• model catalog: https://api.commandcode.ai/provider/v1/models
• pricing documentation: https://commandcode.ai/docs/resources/pricing-limits
• verified: 2026-08-04

The structured pricing payload was independently compared with the 51 live catalog models. All base prices and effective context-tier thresholds matched the committed snapshot.

Notable corrections

• DeepSeek V4 Flash cache read: $0.028/M → $0.0028/M
• Qwen 3.7 Max: remove the expired 50% promotional rates
• MiniMax M2.5: update input/output pricing
• MiMo V2.5 and V2.5 Pro: replace zero costs with current discounted pricing
• explicitly retain zero pricing only for the two models documented as free

Verification

• npm test
• npm run format:check
• npm run typecheck
• git diff --check
• npm pack --dry-run
• real pi extension registration: 51 Command Code models
• independent structured website-payload comparison: 0 pricing mismatches

test:omp-compat skipped locally because omp is not installed on PATH; all other test commands passed.

[github-actions]

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

Scanned Files

None

[github-actions]

Memory benchmark

Compared base 690c640 with PR head 947915d on the same GitHub-hosted darwin runner. Lower values are better.

Metric	Base	PR	PR − Base	Change
Stable RSS	167.3 MiB ± 0.6 MiB	167.9 MiB ± 0.8 MiB	+0.6 MiB ± 1.0 MiB	+0.4%
Physical footprint	110.4 MiB ± 0.6 MiB	111.3 MiB ± 0.8 MiB	+0.5 MiB ± 0.4 MiB	+0.5%
Physical peak	126.3 MiB ± 0.6 MiB	127.6 MiB ± 0.6 MiB	+1.5 MiB ± 1.0 MiB	+1.2%

Interpretation: Differences larger than their measured MAD: Physical footprint +0.5 MiB ± 0.4 MiB, Physical peak +1.5 MiB ± 1.0 MiB.

Extension overhead above pi baseline

Metric	Base overhead	PR overhead	Difference
Stable RSS	+5.3 MiB ± 0.9 MiB	+5.9 MiB ± 1.2 MiB	+0.6 MiB ± 1.0 MiB
Physical footprint	+3.5 MiB ± 0.5 MiB	+4.5 MiB ± 0.8 MiB	+0.5 MiB ± 0.4 MiB
Physical peak	+2.6 MiB ± 0.8 MiB	+3.8 MiB ± 0.9 MiB	+1.5 MiB ± 1.0 MiB

Values are medians of 6 alternating, paired runs. The value after ± is the median absolute deviation (MAD). Each process was sampled 12 times after a 2500 ms warm-up.

Environment: pi 0.82.1, Bun 1.3.11, darwin arm64. RSS comes from ps; physical footprint and peak come from macOS footprint.

This is a comparative signal, not a pass/fail threshold. GitHub-hosted runner noise can affect absolute values.