GLM 5
Z.aiReleased Feb 12, 2026Open weights · MITglm-5
- Intelligence (Epoch)
- 146 · 60th of 111
- Reasoning
- yes
- Knowledge cutoff
- January 2026
- Context
- 203K tokens
- Max output
- 128k tokens
- Input
- text
- API features
- tools, structured outputs, JSON mode, seed, logprobs
- Z.ai's API
- yes, as glm-5
Read Oct 23, 06:00 UTC from the hosts' published offers and OpenRouter, spec sheet from Z.ai · cutoff: aiknowledgecutoff.com
The cheapest host, GMICloud, serves fp8 weights (maker: bf16). Weights and context per host
Intelligence for the price
16 models score higher for the same price or less; it ranks 17 of the 59 models at its price or below.
Epoch Capabilities Index against the lowest input + output price per M, log scale · prices Oct 23. Score: Epoch AI, CC BY 4.0, 146 with an interval of 144 to 148.
Against its rivals
Its previous version, the best of its price range and the top score
- GLM 5this model$2.52146
- Claude Opus 5.5top score$24.00169
- Gemini 3.7 Flashsame price range$4.50158
- Gemini 3.8 Flashsame price range$4.50157
- GLM 4.7previous version$2.15144
Same level for 16× less: Qwen3.7 Flash at Alibaba, $0.16 per M input + output · their Epoch intervals overlap (144 against 146).
Bars: Epoch Capabilities Index (Epoch AI, CC BY 4.0), drawn from 100, below the lowest score in the catalogue · price: lowest input + output per M.
Speed by provider
Output speed of GLM 5 at each host, fastest first
- Amazon Bedrock118 tok/s0.88 s to first token
- GMICloud63 tok/s2.11 s to first token
- StreamLake63 tok/s1.95 s to first token
- Venice62 tok/s1.01 s to first token
- SiliconFlow53 tok/s2.25 s to first token
- Baidu50 tok/s1.00 s to first token
- Z.AI44 tok/s8.15 s to first token
Speed: OpenRouter, median of the last 30 min, read Oct 23, 06:00 UTC · a snapshot of real traffic before the daily read, not a daily average · bars from zero.
- Parameters
- 744B
- 40B active per token (MoE)
- Memory needed
- 446 GB
- in INT4 · 893 GB in INT8
- Smallest cloud setup
- 8× H100
- 640 GB · $17.20/h, about $12,556 a month
- Worth it above
- 5B tokens
- a month, against the cheapest API at $2.52 per M
| Quantization | Memory | Smallest cloud GPU | Quality |
|---|---|---|---|
| INT4 (Q4) | 446 GB | 8× H100 · $17.20/h | smallest footprint, minor quality loss |
| INT8 (Q8) | 893 GB | multi-node, over 640 GB | near-lossless |
| FP16 (full) | 1786 GB | multi-node, over 640 GB | reference quality |
At home: 446 GB in INT4 is beyond any desktop or Mac in our machines table, so it means a multi-GPU server.
Memory ≈ parameters × bytes per parameter × 1.2 · cloud price: on-demand rate, running 24/7 · builds: GGUF, AWQ, MLX, GPTQ, Ollama · break-even calculator
| Provider | Input | Output | Cache ↑ | tok/s | First token | Uptime 24h | Checked |
|---|---|---|---|---|---|---|---|
| GMICloudfp8 (bf16) · 198K ctx−40% | $0.60 | $1.92 | $0.12 | 63 | 2.11 s | 100% | Oct 23 |
| StreamLakefp8 (bf16) · 198K ctx · 125K out−40% | $0.60 | $1.92 | $0.12 | 63 | 1.95 s | 100% | Oct 23 |
| DeepInfrafp4 (bf16) · 198K ctx | $0.60 | $2.08 | $0.12 | — | — | — | Sep 10 |
| Baidufp8 (bf16) · 198K ctx · 128K out | $0.70 | $2.24 | $0.14 | 50 | 1.00 s | 100% | Oct 23 |
| SiliconFlowfp8 (bf16) · 200K ctx · 128K out | $0.95 | $2.55 | $0.20 | 53 | 2.25 s | 99.8% | Oct 23 |
| DigitalOcean64K ctx (200K) | $1.00 | $3.20 | $0.20 | — | — | — | Oct 22 |
| Venicefp8 (bf16) · 198K ctx · 32K out | $1.00 | $3.20 | $0.20 | 62 | 1.01 s | 99.2% | Oct 23 |
| Novita AIfp8 (bf16) · 203K ctx · 128K out | $1.00 | $3.20 | $0.20 | — | — | — | Oct 23 |
| Z.AIfp8 (bf16) · 198K ctx · 128K out | $1.00 | $3.20 | $0.20 | 44 | 8.15 s | 99.2% | Oct 23 |
| Amazon Bedrockprecision not read: host terms · 198K ctx · 128K out | $1.00 | $3.20 | — | 118 | 0.88 s | 100% | Oct 23 |
10 of 10 providers · prices per million tokens, read Oct 23, 06:00 UTC · speed: OpenRouter, median of the last 30 min, read Oct 23, 06:00 UTC · uptime: OpenRouter, last 24 h · −N%: a discount off the host's list price · under each host: weights precision, context and max output as served, the maker's value in brackets when the host's differs
| Provider | Weights | Context | Max output | Tools | Structured outputs | Reasoning |
|---|---|---|---|---|---|---|
| GMICloud | fp8 | 198K | not read: host terms | yes | yes | yes |
| StreamLake | fp8 | 198K | 125K | yes | yes | yes |
| DeepInfra | fp4 | 198K | not stated by the host | — | — | — |
| Baidu | fp8 | 198K | 128K | yes | yes | yes |
| SiliconFlow | fp8 | 200K | 128K | yes | no | yes |
| DigitalOcean | not checked yet | 64K | not checked yet | — | — | — |
| Amazon Bedrock | not read: host terms | 198K | 128K | yes | no | yes |
| Venice | fp8 | 198K | 32K | yes | yes | yes |
| Novita AI | fp8 | 203K | 128K | yes | no | yes |
| Z.AI | fp8 | 198K | 128K | yes | no | yes |
Weights, context and max output: read at each host's own catalogue or docs where its terms allow, otherwise as the host declares to OpenRouter; hover a cell for its source and date · tools, structured outputs, reasoning: as declared to OpenRouter · not verified by us
GLM 5 spec sheet, field by field, with sources
GLM 5 spec sheetmaker-declared
What Z.ai declares about the model itself — read from its own weights repository and documentation, not from any host. What each endpoint actually serves is in the pricing above: the two can differ, and both are true.
- Context window
- 202,752 tokens
- Parameters
- 753,864,139,008≈ 753.9B
- Active experts
- 8 of 256 active per token
- Weight precision
- BF16
- License
- MIT — open weights, permissivecurated · read 2026-09-07Scope: the licence on the published weights, and nothing else. It is not the maker's acceptable-use policy, which is a separate document, and it is not the contract you sign with whichever host you call — hosts set their own terms, which this index does not read.
- Knowledge cutoff
- not published by the maker≈ January 2026 third-party estimate — not the maker's figurecurated
- Modalities
- text
- Max output (official)
- to verify
- Release date
- to verify≈ February 12, 2026 third-party catalogue — not the maker's figure
- Maker lifecycle status
- to verify
Also declared in the weights
- Architecture class
- GlmMoeDsaForCausalLM
- Layers
- 78
- Vocabulary
- 154,880 tokens
- Weights first published
- February 11, 2026
- Weights last updated
- August 11, 2026
- Languages declared
- English · Chinese
- Paper
- arXiv:2602.15763
zai-org/GLM-5 · config.json @ c183ef8 · read 2026-10-08 · 6 of 10 fields published by the maker
| Provider | Chatbot 1,000 in / 500 out | RAG / long context 8,000 in / 500 out | Batch summarize 4,000 in / 1,000 out |
|---|---|---|---|
| GMICloud | $1.56 | $5.76 | $4.32 |
| StreamLake | $1.56 | $5.76 | $4.32 |
| DeepInfra | $1.64 | $5.84 | $4.48 |
| Baidu | $1.82 | $6.72 | $5.04 |
| SiliconFlow | $2.23 | $8.88 | $6.35 |
| DigitalOcean | $2.60 | $9.60 | $7.20 |
| Amazon Bedrock | $2.60 | $9.60 | $7.20 |
| Venice | $2.60 | $9.60 | $7.20 |
| Novita AI | $2.60 | $9.60 | $7.20 |
| Z.AI | $2.60 | $9.60 | $7.20 |
What each host bills for 1,000 requests of a representative job