GLM-5 glm-5
zai-org · glm hybrid-reasoning flagship agentic
unranked — no quality signals yet
- params
- 754B
- arch
- moe
- context
- 200k
- license
- mit open weights
- released
- Feb 2026
- reasoning
- yes
- train compute
- 6.8e+24 FLOP
- downloads/30d
- 1.6M
# architecture
- layers
- 78
- d_model
- 6144
- heads
- 64q / 64kv
- head dim
- 64
- experts
- top-8 of 256 + 1 shared
- vocab
- 154880
- family
- glm_moe_dsa
signal path of one layer, generated from the registry's structured fields — dimension lines quote the model's real numbers; an MoE trace forks at the router, a dense trace runs straight through. Geometry fields come from the repo's config.json.
# why ranked
Overall: provisional — score 68.7 ● low — a single signal; treat with caution
provisional: too few independent signals for a numbered position — the score is shown, but this model sorts after every ranked model.
| signal | weight | input (0–100) |
|---|---|---|
| aa_intelligence | 0.70 | — |
| bench_composite | 0.30 | 68.7 |
benchmark panel evidence:
| complete panel | score (0–100) |
|---|---|
| browsecomp-v1 | 51.9 |
| swe-bench-verified-v1 | 85.4 |
missing signals are dropped and the remaining weights renormalized — never imputed.
Coding: provisional — score 94.7 ● low — a single signal; treat with caution
provisional: too few independent signals for a numbered position — the score is shown, but this model sorts after every ranked model.
| signal | weight | input (0–100) |
|---|---|---|
| aa_coding | 0.30 | — |
| aider_polyglot | 0.30 | — |
| swe_bench_verified | 0.40 | 94.7 |
missing signals are dropped and the remaining weights renormalized — never imputed.
Agentic: #3 — score 94.7 ● low — a single signal; treat with caution
| signal | weight | input (0–100) |
|---|---|---|
| swe_bench_verified | 0.60 | 94.7 |
| swe_rebench | 0.40 | — |
missing signals are dropped and the remaining weights renormalized — never imputed.
# trend
methodology changed during this history window; score movement across that boundary is not model movement. See methodology v6.
# benchmarks
# where to run
| provider | quant | ctx | $/M in | $/M out | $/M cache | price src | tps | uptime | ✓ |
|---|---|---|---|---|---|---|---|---|---|
| Featherless | — | — | — | — | via hfrouter | — | — | ||
| GMICloud | fp8 | 198k | $0.60 | $1.92 | $0.12 | via openrouter | — | 99.7% | |
| StreamLake | fp8 | 198k | $0.60 | $1.92 | $0.12 | via openrouter | — | 100.0% | |
| DeepInfra | 198k | $0.60 | $2.08 | — | via hfrouter | 24 | — | ||
| Baidu | fp8 | 198k | $0.70 | $2.24 | $0.14 | via openrouter | — | 100.0% | |
| SiliconFlow | fp8 | 200k | $0.95 | $2.55 | $0.20 | openrouter | — | 99.8% | ✓ |
| AtlasCloud | fp8 | 198k | $0.95 | $3.15 | — | atlascloud | — | — | ✓ |
| BaseTen | — | $0.95 | $3.15 | — | via litellm | — | — | ||
| Amazon Bedrock | unknown | 198k | $1.00 | $3.20 | — | bedrock | — | 100.0% | ✓ |
| Novita | fp8 | 202k | $1.00 | $3.20 | — | novita | — | 100.0% | ✓ |
| Together | 198k | $1.00 | $3.20 | — | via modelsdev | — | — | ||
| Venice | fp8 | 198k | $1.00 | $3.20 | — | venice | — | 100.0% | ✓ |
| Venice | unknown | 200k | $1.00 | $3.20 | — | venice | — | 100.0% | ✓ |
| Z.AI | fp8 | 198k | $1.00 | $3.20 | $0.20 | via openrouter | — | 100.0% |
sorted by blended price ((3·input + output) / 4 per 1M) · ✓ = the provider's own catalog confirms the offer · "via …" prices are what the aggregator routing the offer charges, not the provider's own list price
# variants
source aliases
- aa
glm-5- epoch
GLM-5- openrouter
z-ai/glm-5