codex-usage v2.7.1
新增 GPT-6 Sol 和 GPT-6 Luna 的内置定价与 Codex Fast 费用折算。升级后，这两个模型已记录的用量可以直接参与 API 等价成本估算，无需手动配置模型映射。
| 模型 | Input | Cached Input | Cache Write | Output |
|---|---|---|---|---|
| GPT-6 Sol | 2.00 | 0.20 | 2.50 | 10.00 |
| GPT-6 Luna | 0.10 | 0.01 | 0.125 | 0.50 |
以上为 2026-09-23 核对的 Standard 短上下文文本价格，单位 USD / 1M Token。
- 支持正式模型标识
gpt-6-sol、gpt-6-luna，沿用有效日期后缀的快照匹配规则，并可作为本机内部模型映射的目标。 - 两个模型明确标记的 Fast 用量按 ChatGPT Codex 2.5× 额度倍率折算；原始 Token 数量保持不变。该估算与 API Fast 计费口径不同，仍不推断长上下文附加费。
- Dashboard 内置价格表、模型映射选项、synthetic Demo 和中英文文档同步更新。
- 沿用 schema v11，从 v2.7.0 升级无需为此次模型目录更新重建历史。原有模型单价与用户已有的内部模型映射保持不变；未知模型继续标为未定价。
验证覆盖四类 Token 单价、缓存写入精度、快照匹配、内部模型映射、Fast 加权和真实服务的 JSONL → API → Dashboard 流程，并运行 Go 测试、静态检查及 Dashboard / synthetic Demo 浏览器回归。
Add built-in pricing and Codex Fast cost estimates for GPT-6 Sol and GPT-6 Luna. Existing recorded usage for these models becomes priceable after upgrading without a manual model mapping.
Standard short-context text rates verified on September 23, 2026, in USD / 1M tokens:
| Model | Input | Cached Input | Cache Write | Output |
|---|---|---|---|---|
| GPT-6 Sol | 2.00 | 0.20 | 2.50 | 10.00 |
| GPT-6 Luna | 0.10 | 0.01 | 0.125 | 0.50 |
- Resolve the exact model IDs and valid dated snapshots; allow explicit internal-model aliases to either model.
- Apply the documented ChatGPT Codex 2.5× credit multiplier to confirmed Fast usage without changing raw token counts. This remains distinct from API Fast billing and does not infer long-context surcharges.
- Update the Dashboard catalog, alias choices, synthetic Demo, and Chinese/English documentation.
- Retain schema v11. No history rebuild is needed for this catalog update when upgrading from v2.7.0. Existing model rates and internal-model mappings are retained; unknown models remain unpriced.
Validation covers token-category rates, cache-write precision, snapshot matching, aliases, Fast weighting, and the real-service JSONL → API → Dashboard path, plus Go tests, vet, and Dashboard / synthetic Demo browser regressions.