GLM-5 glm-5
zai-org · glm hybrid-reasoning flagship agentic
#11 overall
- params
- 754B
- arch
- moe
- context
- 198k
- license
- mit open weights
- released
- Feb 2026
- reasoning
- yes
- train compute
- 6.8e+24 FLOP
- downloads/30d
- 70.0k
# why ranked
Overall: #11 — score 24.7 ● medium — one signal missing, weights renormalized
| signal | weight | input (0–100) |
|---|---|---|
| aa_intelligence | 0.50 | 39.5 |
| arena_elo | 0.30 | 0.0 |
| bench_composite | 0.20 | — |
missing signals are dropped and the remaining weights renormalized — never imputed.
# trend
# benchmarks
| benchmark | score | source | date |
|---|---|---|---|
| AA Intelligence Index | 39.5 | Artificial Analysis | — |
| LMArena Elo | 1457.0 | LMArena | Jul 2026 |
# where to run
| provider | quant | ctx | $/M in | $/M out | $/M blend | tps | uptime |
|---|---|---|---|---|---|---|---|
| GMICloud | fp8 | 198k | $0.60 | $1.92 | $0.93 | — | 99.7% |
| DeepInfra | fp4 | 198k | $0.60 | $2.08 | $0.97 | — | 100.0% |
| StreamLake | fp8 | 198k | $0.65 | $2.08 | $1.01 | — | 99.9% |
| Baidu | fp8 | 198k | $0.70 | $2.24 | $1.08 | — | 100.0% |
| DigitalOcean | unknown | 64k | $0.75 | $2.40 | $1.16 | — | 95.9% |
| Chutes | fp8 | 198k | $0.95 | $2.55 | $1.35 | — | — |
| SiliconFlow | fp8 | 200k | $0.95 | $2.55 | $1.35 | — | 100.0% |
| AtlasCloud | fp8 | 198k | $0.95 | $3.15 | $1.50 | — | 100.0% |
| Amazon Bedrock | unknown | 198k | $1.00 | $3.20 | $1.55 | — | 96.1% |
| Novita | fp8 | 202k | $1.00 | $3.20 | $1.55 | — | 100.0% |
| Parasail | fp8 | 198k | $1.00 | $3.20 | $1.55 | — | 98.5% |
| Venice | fp8 | 198k | $1.00 | $3.20 | $1.55 | — | — |
| Z.AI | fp8 | 198k | $1.00 | $3.20 | $1.55 | — | 93.0% |
| Phala | unknown | 198k | $1.20 | $3.50 | $1.77 | — | — |
blended = (3·input + output) / 4 per 1M tokens · cheapest first
# variants
source aliases
- aa
glm-5- epoch
GLM-5- openrouter
z-ai/glm-5