Skip to content

Commit 8a87b43

Browse files
鲁工鲁工
authored andcommitted
feat: add GLM-5.3-Flash and Qwen3.8 Flash, v1.16.0
Two new multimodal flash-tier models, added as variants on the existing providers so no new keys or endpoints are introduced; defaults unchanged. GLM-5.3-Flash (glm-plan-5.3-flash, glm-global-5.3-flash): first native multimodal GLM-5 model, 1M context / 128K max output, thinking cannot be disabled. Both vendor Claude Code guides (docs.bigmodel.cn coding-plan/ tool/claude, docs.z.ai devpack/tool/claude) configure glm-5.3-flash on the same Anthropic endpoints already used by glm-plan / glm-global. Specs: docs.bigmodel.cn/cn/guide/models/vlm/glm-5.3-flash and docs.z.ai/guides/vlm/glm-5.3-flash. Unverified upstream: no GLM keys. Qwen3.8 Flash (qwen3.8-flash, qwen-plan-3.8-flash): 1M context / 131,072 max output, image+video input; listed in Bailian's Anthropic- compatible Messages API model set (help.aliyun.com/zh/model-studio/ anthropic-api-messages). Pay-go verified: `ccmr doctor qwen3.8-flash` -> [OK] 1.69s. Token Plan variant kept on the strength of the platform.qianwenai.com latest-model doc ("Token Plan supports Qwen3.8-Flash"); the overview tier tables do not list it but are stale (still show retired glm-5.1). Not verified live: the test subscription returns 403 AccessDenied.Unpurchased for the shipped 3.8-max as well. qwen3.8-flash-next is deliberately NOT added: it is the open-weights Qwen4-architecture preview (HF Qwen/Qwen3.8-Flash-Next); Bailian 404s on it and no first-party catalog or OpenRouter serves that id; the HF card names qwen3.8-flash as the production API. A test pins its absence. The international Qwen line (dashscope-intl / QwenCloud) exists but is out of scope by user decision.
1 parent 7bba049 commit 8a87b43

8 files changed

Lines changed: 307 additions & 3 deletions

File tree

README.md

Lines changed: 18 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -200,12 +200,16 @@ ccmr claude --print --output-format json "你的问题"
200200
| `minimax-global-m3` | `minimax-global`, `minimax-io` | MiniMax M3 | MiniMax Global |
201201
| `qwen3.8-max` | `qwen`, `tongyi`, `qwen3.8` | Qwen3.8 Max | 阿里云(按量付费) |
202202
| `qwen3.7-max` | `qwen-max`, `qwen3.7` | Qwen3.7 Max | 阿里云(按量付费) |
203+
| `qwen3.8-flash` | `qwen-flash` | Qwen3.8 Flash(多模态,支持图片/视频输入) | 阿里云(按量付费) |
203204
| `qwen-plan-3.8-max` | `qwen-plan`, `qwen-plan-3.8`, `qwen-plan-max` | Qwen3.8 Max | 千问 Token Plan(订阅) |
204205
| `qwen-plan-3.7-max` | `qwen-plan-3.7` | Qwen3.7 Max | 千问 Token Plan(订阅) |
206+
| `qwen-plan-3.8-flash` | `qwen-plan-flash` | Qwen3.8 Flash(多模态,支持图片/视频输入) | 千问 Token Plan(订阅) |
205207
| `glm-plan-5.3` | `glm-plan`, `glm`, `glm-5.3`, `zhipu`, `chatglm` | GLM-5.3 | 智谱 GLM Coding Plan(订阅) |
206208
| `glm-plan-5.2` | `glm-5.2` | GLM-5.2 | 智谱 GLM Coding Plan(订阅) |
209+
| `glm-plan-5.3-flash` | `glm-flash`, `glm-5.3-flash`, `glm-plan-flash` | GLM-5.3-Flash(原生多模态) | 智谱 GLM Coding Plan(订阅) |
207210
| `glm-global-5.3` | `glm-global`, `zai`, `z-ai` | GLM-5.3 | Z.ai(国际) |
208211
| `glm-global-5.2` | - | GLM-5.2 | Z.ai(国际) |
212+
| `glm-global-5.3-flash` | `glm-global-flash`, `zai-flash` | GLM-5.3-Flash(原生多模态) | Z.ai(国际) |
209213
| `step-3.7-flash` | `step`, `step-3.7`, `stepfun` | Step 3.7 Flash | 阶跃星辰(按量付费) |
210214
| `step-plan-3.7-flash` | `step-plan`, `step-plan-3.7`, `stepplan` | Step 3.7 Flash (Step Plan) | 阶跃星辰(订阅) |
211215
| `mimo-v2.5-pro` | `mimo`, `mimo-pro`, `mimo-token-sgp`, `xiaomi` | MiMo V2.5 Pro | MiMo Token Plan SGP |
@@ -233,7 +237,9 @@ ccmr claude --print --output-format json "你的问题"
233237
| MiniMax M3 (CN / Global) | 1M | 128K |
234238
| Qwen3.8 Max (按量付费 / Token Plan) | 1M | 64K |
235239
| Qwen3.7 Max (按量付费 / Token Plan) | 1M | 64K |
240+
| Qwen3.8 Flash (按量付费 / Token Plan) | 1M | 128K |
236241
| GLM-5.3 / GLM-5.2 (Coding Plan / 国际) | 1M | 128K |
242+
| GLM-5.3-Flash (Coding Plan / 国际) | 1M | 128K |
237243
| Step 3.7 Flash (按量付费 / Step Plan) | 256K | 384K |
238244
| MiMo V2.5 Pro | 1M | 128K |
239245
| MiMo V2.5 | 1M | 128K |
@@ -415,6 +421,8 @@ npx claude-code-model-router claude
415421
/model glm # 切换到 GLM-5.3(智谱 Coding Plan 订阅)
416422
/model glm-global # 切换到 GLM-5.3(国际 Z.ai)
417423
/model deepseek-vision # 切换到 DeepSeek V4 Flash Vision(实验版,支持图片输入)
424+
/model qwen-flash # 切换到 Qwen3.8 Flash(按量付费,多模态)
425+
/model glm-flash # 切换到 GLM-5.3-Flash(智谱 Coding Plan 订阅,原生多模态)
418426
/model seed # 切换到 Doubao Seed 2.1 Pro(火山方舟 按量付费)
419427
/model seed-plan # 切换到 Doubao Seed 2.1 Pro(火山方舟 Agent Plan 订阅)
420428
/model step # 切换到 Step 3.7 Flash(按量付费)
@@ -433,8 +441,10 @@ npx claude-code-model-router claude
433441
/model deepseek-v4-flash # DeepSeek V4 Flash
434442
/model glm-plan-5.3 # GLM-5.3(智谱 Coding Plan 订阅)
435443
/model glm-plan-5.2 # GLM-5.2(智谱 Coding Plan 订阅)
444+
/model glm-plan-5.3-flash # GLM-5.3-Flash(智谱 Coding Plan 订阅;别名 glm-flash)
436445
/model glm-global-5.3 # GLM-5.3(国际 Z.ai)
437446
/model glm-global-5.2 # GLM-5.2(国际 Z.ai)
447+
/model glm-global-5.3-flash # GLM-5.3-Flash(国际 Z.ai;别名 glm-global-flash)
438448
/model step-3.7-flash # Step 3.7 Flash(按量付费)
439449
/model step-plan-3.7-flash # Step 3.7 Flash(Step Plan 订阅)
440450
/model minimax-m3 # MiniMax M3
@@ -446,6 +456,8 @@ npx claude-code-model-router claude
446456
/model qwen3.7-max # Qwen3.7 Max(按量付费)
447457
/model qwen-plan-3.8-max # Qwen3.8 Max(Token Plan 订阅)
448458
/model qwen-plan-3.7-max # Qwen3.7 Max(Token Plan 订阅)
459+
/model qwen3.8-flash # Qwen3.8 Flash(按量付费;别名 qwen-flash)
460+
/model qwen-plan-3.8-flash # Qwen3.8 Flash(Token Plan 订阅;别名 qwen-plan-flash)
449461
/model seed-2.1-pro # Doubao Seed 2.1 Pro(按量付费)
450462
/model seed-2.1-turbo # Doubao Seed 2.1 Turbo(按量付费)
451463
/model seed-plan-2.1-pro # Doubao Seed 2.1 Pro(Agent Plan 订阅)
@@ -533,6 +545,12 @@ DeepSeek Anthropic 兼容接口会忽略 `metadata` 字段,但某些 Claude Co
533545

534546
## 更新日志
535547

548+
### v1.16.0
549+
550+
- **新增 GLM-5.3-Flash**[智谱文档](https://docs.bigmodel.cn/cn/guide/models/vlm/glm-5.3-flash) / [Z.ai 文档](https://docs.z.ai/guides/vlm/glm-5.3-flash)):GLM-5 系列首个原生多模态模型(视频/图片/文本/文件输入),320B 总参、18B 激活,1M 上下文、最大输出 128K,定价为 GLM-5.3 的 1/10。国内 `glm-plan-5.3-flash`(别名 `glm-flash` / `glm-5.3-flash` / `glm-plan-flash`)与国际 `glm-global-5.3-flash`(别名 `glm-global-flash` / `zai-flash`)同步新增,端点与 Key 复用现有 provider;两份官方 Claude Code 配置均已直接使用 `glm-5.3-flash`。Coding Plan 额度为 GLM-5.3 的 3 倍。默认模型不变(`glm` / `glm-global` 仍指向 5.3)。与 GLM-5.3 一样强制开启思考
551+
- **新增 Qwen3.8 Flash**([百炼模型信息](https://help.aliyun.com/zh/model-studio/qwen3-8-flash) / [百炼 Anthropic 兼容 API](https://help.aliyun.com/zh/model-studio/anthropic-api-messages)):千问多模态 flash 档(图片/视频/文本输入),1M 上下文、最大输出 131,072;百炼北京按量价 ¥0.8 / ¥2.7 每百万 tokens(输入/输出)。按量 `qwen3.8-flash`(别名 `qwen-flash`)与 Token Plan `qwen-plan-3.8-flash`(别名 `qwen-plan-flash`)同步新增;默认模型不变(`qwen` / `qwen-plan` 仍指向 3.8 Max)。按量版实测 `ccmr doctor` 通过;Token Plan 版依据[千问 AI 平台 latest-model 文档](https://platform.qianwenai.com/docs/developer-guides/getting-started/latest-model)「Token Plan 支持 Qwen3.8-Flash」收录,但 [Token Plan 概述](https://platform.qianwenai.com/docs/token-plan/overview)的档位模型表尚未列出它(该表仍含已下线的 glm-5.1,应为过期),且本次未能实测(测试订阅对 3.8-max 同样返回 `403 AccessDenied.Unpurchased`)——若你的档位不含该模型,上游会返回同样的 403。国际版:百炼国际站(`{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com`)与 QwenCloud(`dashscope-intl`)同样提供该模型,但 ccmr 目前没有国际版 Qwen provider,需要的话请开 issue
552+
- **关于 `qwen3.8-flash-next`**:它是 Qwen4 架构的开源预览版权重([HF](https://huggingface.co/Qwen/Qwen3.8-Flash-Next),Qwen Community License),百炼 / 千问 AI 平台 / QwenCloud / OpenRouter 均无该 id 的托管 API,官方说明线上正式版即 `qwen3.8-flash`。因此不提供路由——需要它的话直接用 `qwen3.8-flash`
553+
536554
### v1.15.0
537555

538556
- **新增 GLM-5.3**[智谱文档](https://docs.bigmodel.cn/cn/guide/models/text/glm-5.3) / [Z.ai 文档](https://docs.z.ai/guides/llm/glm-5.3)):智谱最新旗舰,1M 上下文、最大输出 128K,GLM Coding Plan 已全量上线。国内 `glm-plan-5.3` 与国际 `glm-global-5.3` 同步新增并设为默认——`glm` / `zhipu` / `glm-plan` 现指向 5.3,`glm-global` / `zai` 同理;5.2 保留(`glm-5.2` / `glm-plan-5.2` / `glm-global-5.2`)。注意:GLM-5.3 强制开启思考(`reasoning_effort` 支持 low/high/max,默认 max),上游不再接受 `thinking.type: "disabled"`

dist/config.d.ts.map

Lines changed: 1 addition & 1 deletion
Some generated files are not rendered by default. Learn more about customizing how changed files appear on GitHub.

dist/config.js

Lines changed: 90 additions & 0 deletions
Some generated files are not rendered by default. Learn more about customizing how changed files appear on GitHub.

dist/config.js.map

Lines changed: 1 addition & 1 deletion
Some generated files are not rendered by default. Learn more about customizing how changed files appear on GitHub.

0 commit comments

Comments
 (0)