GLM-5.2 auf OpenRouter gesichtet
Offene Modellgewichte28 Beleglinks
Englisches Original

Z.ai
USD / 1 Mio. Token
| Eingabe | Ausgabe | Cache-Eingabe |
|---|---|---|
|
|
|
| Modell | Beleglinks | Preis Ein-/Ausgabe |
|---|---|---|
glm-5.2standard · Text | opencode-go | $1,40 / $4,40 |
glm-5.2standard | zai-pricing API-Route nicht verfügbar | $1,40 / $4,40 |
gain. Models that run top-2048 over every key (GLM-5.2/5.3) should gain more; that is not measured here.
glm-agentx branch (#1809, #2107, #2201), that brings the GLM-5.2 decode optimizations to main. Each kernel comes first in its own PR, and a separate PR
glm-agentx branch (#1809, #2107, #2201), that brings the GLM-5.2 decode optimizations to main. Each kernel comes first in its own PR, and a separate PR
glm-agentx branch (#1809, #2107, #2201), that brings the GLM-5.2 decode optimizations to main. Each kernel comes first in its own PR, and a separate PR
glm-agentx branch (#1809, #2107, #2201), that brings the GLM-5.2 decode optimizations to main. Each kernel comes first in its own PR, and a separate PR
glm-agentx branch (#1809, #2107, #2201), that brings the GLM-5.2 decode optimizations to main. Each kernel comes first in its own PR, and a separate PR
glm-agentx branch (#1809, #2107, #2201), that brings the GLM-5.2 decode optimizations to main. Each kernel comes first in its own PR, and a separate PR
glm-agentx branch (#1809, #2107, #2201), that brings the GLM-5.2 decode optimizations to main. Each kernel comes first in its own PR, and a separate PR
glm-agentx branch (#1809, #2107, #2201), that brings the GLM-5.2 decode optimizations to main. Each kernel comes first in its own PR, and a separate PR
At decode, every decoder layer of GLM-5.2 ends with a TP all-reduce, which sums the layer's output across the 16 cores. This tensor is small.
- model `z-ai/glm-5.2`;
| `nano-gpt/z-ai/glm-5.2` | link → `zhipu/glm-5.2` | NanoGPT's GLM 5.2 is Z.ai's flagship 1M-context coding model released June 2026 (NVIDIA NIM, Vercel), matching zhipu/glm-5.2 (2026-06-13,
only for models that accept it (qwen3.8-max, kimi-k3, glm-5.2) and stripped elsewhere to avoid vendor 400s; `max` preserved on kimi-k3. Since opening, live dogfooding corrected
- updated: `/home/runner/_work/models.dev/models.dev/providers/nano-gpt/models/z-ai/glm-5.2:thinking.toml`
- Files per model: qwen3.5 160, glm-5.2 147, kimi-k3 119, qwen3-next 80, deepseek-v4-pro 53, glm-5.3 53, deepseek-v4-flash 45, deepseek-r1 40, deepseek-v3 40, minimax-m2.5 32, muse-spark
the 11 ids that /v1/models no longer publishes (GLM-5.2, DeepSeek
GLM-5.2 ist ein Modell von Z.ai mit offenen Modellgewichten, veröffentlicht am 2026-06-16. Das Kontextfenster umfasst 1.048.576 Tokens. Die maximale Ausgabelänge beträgt 131.072 Tokens. Pro Million Tokens kostet die Eingabe $1,40 und die Ausgabe $4,40.
Das Kontextfenster umfasst 1.048.576 Tokens. Die maximale Ausgabelänge beträgt 131.072 Tokens.
Pro Million Tokens kostet die Eingabe $1,40 und die Ausgabe $4,40. Cache-Eingabe kostet $0,26 pro Million.
Ja. Z.ai veröffentlicht die Modellgewichte.
Am 2026-06-16.