Models
平台托管的模型服务与网关。开通模型、查看用量、配置路由规则；开通后的模型可在 Model Configuration 中引用。
Total Requests
12,847
Total Cost
$284.60
Total Tokens
96.4M
Models Performance
FIELDS
| Model | Requests | Tokens | Cost | Avg Latency | P95 | Error Rate |
|---|---|---|---|---|---|---|
| claude-sonnet-5 | 8,204 | 61.2M | $196.40 | 3.2s | 8.1s | 0.4% |
| claude-haiku-4-5 | 3,102 | 18.6M | $41.20 | 1.8s | 4.2s | 0.2% |
| deepseek-v4 | 1,398 | 14.1M | $38.90 | 2.4s | 6.0s | 0.9% |
| gpt-4o | 143 | 2.5M | $8.10 | 2.9s | 7.3s | 1.6% |
用量按 Model Gateway 的实际调用采集；Experiment 与 Service 产生的调用均计入。
FIELDS
| Name | Enable | Price $/1M | Context | Max Output | Capabilities | Intelligence | Status | Released |
|---|---|---|---|---|---|---|---|---|
| Claude Sonnet 5anthropic/claude-sonnet-5 | ↗$3.00↙$15.00 | 200K | 65,536 | 71 | Available | 2026-08-13 | ||
| Claude Haiku 4.5anthropic/claude-haiku-4-5 | ↗$0.80↙$4.00 | 200K | 32,768 | 55 | Available | 2026-08-13 | ||
| GPT-5.2openai/gpt-5.2 | ↗$2.50↙$10.00 | 256K | 65,536 | 69 | Available | 2026-08-13 | ||
| o4-miniopenai/o4-mini | ↗$1.10↙$4.40 | 128K | 32,768 | 62 | Available | 2026-08-13 | ||
| DeepSeek V4deepseek/deepseek-v4 | ↗$0.27↙$1.10 | 128K | 32,768 | 58 | Available | 2026-08-13 | ||
| DeepSeek R2deepseek/deepseek-r2 | ↗$0.55↙$2.19 | 128K | 65,536 | 60 | Preview | 2026-08-13 | ||
| Qwen3-Maxqwen/qwen3-max | ↗$1.60↙$6.40 | 256K | 32,768 | 53 | Available | 2026-08-13 | ||
| GPT-4oopenai/gpt-4o | ↗$2.50↙$10.00 | 128K | 16,384 | 41 | Deprecated | 2026-08-13 | ||
| text-embedding-4openai/text-embedding-4 | ↗$0.10↙— | 8K | — | — | Available | 2026-08-13 | ||
| qwen3-embeddingqwen/qwen3-embedding | ↗$0.07↙— | 32K | — | — | Available | 2026-08-13 | ||
| Kimi K2.5moonshot/kimi-k2-5 | ↗$0.60↙$2.50 | 256K | 32,768 | 57 | Available | 2026-07-30 | ||
| Kimi K2moonshot/kimi-k2 | ↗$0.55↙$2.20 | 128K | 16,384 | 50 | Available | 2026-03-18 | ||
| Claude Opus 4.5anthropic/claude-opus-4-5 | ↗$5.00↙$25.00 | 200K | 65,536 | 73 | Available | 2026-06-11 | ||
| GPT-5.2 miniopenai/gpt-5.2-mini | ↗$0.45↙$1.80 | 256K | 32,768 | 60 | Available | 2026-08-13 | ||
| Qwen3-Coderqwen/qwen3-coder | ↗$0.90↙$3.60 | 128K | 32,768 | 56 | Preview | 2026-08-02 | ||
| DeepSeek V3.2deepseek/deepseek-v3.2 | ↗$0.14↙$0.55 | 64K | 8,192 | 44 | Deprecated | 2025-12-09 | ||
| GPT-Image 2openai/gpt-image-2 | ↗$8.00↙— | 32K | — | — | Available | 2026-05-27 | ||
| Qwen-Image-Plusqwen/qwen-image-plus | ↗$2.40↙— | 16K | — | — | Available | 2026-07-15 | ||
| Whisper Large v4openai/whisper-large-v4 | ↗$0.35↙— | — | — | — | Available | 2026-04-22 | ||
| Qwen3-TTSqwen/qwen3-tts | ↗$1.20↙— | 8K | — | — | Preview | 2026-08-20 |
没有匹配的模型
清除筛选试试 — Clear filters
模型开通状态属于当前 Workspace。已开通模型的月度配额管理将在后续版本提供；Deprecated 模型可继续使用至退役日期。Intelligence 取自第三方 Artificial Analysis Intelligence Index（2026-08 快照，满分 100），随上游榜单更新；Embedding 模型不在该 index 覆盖范围内，显示为 —。
Routing Rules 3
按 Priority 顺序匹配，首条命中生效；仅作用于 Service 生产流量。
| Priority | Rule Name | Strategy | Match | Target Chain | Enabled | Actions |
|---|---|---|---|---|---|---|
| P1 | 旗舰降级链 | Fallback | All Requests | claude-sonnet-5→claude-haiku-4-5 Trigger: Timeout / 5xx |
||
| P2 | 成本分流 | Load Balance | config: support-agent-model | deepseek-v4(60%) + claude-haiku-4-5(40%) | ||
| P3 | GPT-4o 迁移 | Fallback | config: legacy-gpt4o | gpt-4o→gpt-5.2 |
路由规则作用于 Model Gateway 层的 Service 生产流量；Experiment 产生的 Run 默认绕过路由规则，严格按 Revision 的 Composition 直连声明的模型——否则单变量对照的前提会被静默破坏。规则按 Priority 顺序匹配，首条命中生效；生产流量命中切换时，追溯链记录实际调用的模型。