Zhipu AI · China
GLM-5
glm-5
NovedadOpen-weightMás barato
Familia GLM. Top en coding agéntico open-weight.
Input / 1M
$1
USD
Output / 1M
$3.2
USD
Contexto
205K
tokens
SWE-bench
77.8%
Pro / Verified
Fortalezas
- SWE-bench alto: encaja en tareas reales de ingeniería.
- Precio de entrada muy competitivo por millón de tokens.
- Open-weight: puedes hospedarlo y auditar los pesos.
- Pensado para flujos agénticos de coding.
Matices
Casos de uso
- Agentes de código y tareas de ingeniería de software.
- Prototipos, CI y volumen alto con presupuesto controlado.
- Despliegue local o en infraestructura propia.
Límites y ficha
- Contexto
- 205K
- Modalidades
- text
- Licencia
- Open-weight
- Parámetros
- —
- Lanzamiento
- abr 2026
- Origen
- China
Benchmarks
- SWE-bench
- 77.8%
- LiveCodeBench
- —
- HumanEval
- —
Notas
~77,8% en SWE-bench Verified: top open-weight en coding agéntico. Precio aproximado. Competidor directo de Kimi K2.6.
Historial de la ficha
- 2026-09-29Ficha refrescada con precios, contexto y benchmarks del Intelligence Index.
Actualizado en catálogo el 29 sept 2026. Confirma precio y límites en la web oficial.