GLM-5.3-Flash
Z.aiReleased Aug 26, 2026Open weights · MITglm-5.3-flash
- Intelligence (Epoch)
- 153 · 34th of 111
- Reasoning
- yes, with effort levels
- Knowledge cutoff
- January 2026
- Context
- 1M tokens
- Max output
- 128K tokens
- Input
- text, image, video
- API features
- tools, structured outputs, JSON mode, seed, logprobs
- Z.ai's API
- yes, as glm-5.3-flash
Read Oct 23, 06:00 UTC from the hosts' published offers and OpenRouter, spec sheet from Z.ai · cutoff: aiknowledgecutoff.com
Lowest input, per M$0.0435at StreamLake
Cheapest cache, per M$0.0075at Relace
Fastest140 tok/sat CoreWeave, on OpenRouter
Providers29serve this version
Intelligence for the price
On the best-deal line: no model scores higher for less than $0.19 per M.
Epoch Capabilities Index against the lowest input + output price per M, log scale · prices Oct 23. Score: Epoch AI, CC BY 4.0, 153 with an interval of 151 to 156.
Against its rivals
The best of its price range and the top score
- GLM-5.3-Flashthis model$0.19153
- Claude Opus 5.5top score$24.00169
- DeepSeek V4 Flash 0731same price range$0.21155
- DeepSeek V4 Flash 0423same price range$0.21146
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-5.3-Flash at each host, fastest first (the 10 fastest of 23)
- CoreWeave140 tok/s0.57 s to first token
- DigitalOcean132 tok/s0.81 s to first token
- BaseTen122 tok/s0.47 s to first token
- Together AI98 tok/s5.12 s to first token
- Fireworks AI81 tok/s1.04 s to first token
- InferenceNet79 tok/s0.45 s to first token
- Crusoe79 tok/s0.90 s to first token
- Parasail76 tok/s1.36 s to first token
- Inceptron75 tok/s0.41 s to first token
- Friendli68 tok/s4.42 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
- 320B
- 18B active per token (MoE)
- Memory needed
- 192 GB
- in INT4 · 384 GB in INT8
- Smallest cloud setup
- 4× H100
- 320 GB · $8.60/h, about $6,278 a month
- Worth it above
- 33B tokens
- a month, against the cheapest API at $0.19 per M
| Quantization | Memory | Smallest cloud GPU | Quality |
|---|---|---|---|
| INT4 (Q4) | 192 GB | 4× H100 · $8.60/h | smallest footprint, minor quality loss |
| INT8 (Q8) | 384 GB | 8× H100 · $17.20/h | near-lossless |
| FP16 (full) | 768 GB | multi-node, over 640 GB | reference quality |
At home: in INT4 it fits a Mac Studio M5 Ultra (256 GB unified), about $9,499 once, or $0.3615/h over 3 years, electricity aside; it pays for itself after about 50B tokens against the cheapest API.
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 |
|---|---|---|---|---|---|---|---|
| Relaceprecision not stated · 1M ctx · 128K outreprices often | $0.075 | $0.50 | $0.0075 | 56 | 1.16 s | 100% | Oct 23 |
| StreamLakefp8 · 1000K ctx · 125K out−71%reprices often | $0.0435 | $0.14 | $0.0087 | 28 | 2.65 s | 78.9% | Oct 23 |
| Pareto Inferenceprecision not stated · 128K out | $0.06 | $0.20 | $0.012 | — | — | — | Oct 23 |
| GMICloudfp8 · 1M ctx−50% | $0.075 | $0.25 | $0.015 | 27 | 4.79 s | 94.7% | Oct 23 |
| CoreWeavenvfp4 (fp8) · 1M ctx | $0.15 | $0.50 | $0.02 | 140 | 0.57 s | 99.0% | Oct 23 |
| Io Netprecision not read: host terms · 256K ctx (1M) · 64K out−25% | $0.11 | $0.38 | $0.0225 | — | — | 89.7% | Oct 4 |
| Phalanvfp4 (fp8) · 1M ctx · 128K out−15% | $0.13 | $0.42 | $0.0255 | 20 | 4.96 s | 98.9% | Oct 23 |
| InferenceNetfp4 (fp8) · 1M ctx · 256K out | $0.10 | $0.45 | $0.03 | 79 | 0.45 s | 100% | Oct 23 |
| Novita AIfp8 · 1M ctx · 128K out | $0.15 | $0.50 | $0.03 | — | — | — | Oct 23 |
| DeepInfrafp4 (fp8) · 1M ctx | $0.15 | $0.50 | $0.03 | 15 | 1.30 s | 99.9% | Oct 23 |
| Z.AIfp8 · 1M ctx · 128K out | $0.15 | $0.50 | $0.03 | 40 | 4.13 s | 100% | Oct 23 |
| Modalnvfp4 (fp8) · 1M ctx | $0.15 | $0.50 | $0.03 | — | — | — | Oct 23 |
| Fireworks AI1M ctx | $0.15 | $0.50 | $0.03 | 81 | 1.04 s | 96.2% | Oct 23 |
| Friendliprecision not read: host terms · 1M ctx | $0.15 | $0.50 | $0.03 | 68 | 4.42 s | 100% | Oct 23 |
| SiliconFlowfp8 · 1M ctx · 256K out | $0.15 | $0.50 | $0.03 | 28 | 2.85 s | 83.9% | Oct 23 |
| DigitalOcean1M ctx · 128K out | $0.15 | $0.50 | $0.03 | 132 | 0.81 s | 100% | Oct 23 |
| Together AIfp4 (fp8) · 1.05M ctx | $0.15 | $0.50 | $0.03 | 98 | 5.12 s | 91.7% | Oct 23 |
| Parasailfp4 (fp8) · 1M ctx | $0.15 | $0.50 | $0.03 | 76 | 1.36 s | 100% | Oct 23 |
| BaseTenfp8 · 1M ctx · 128K out | $0.15 | $0.50 | $0.03 | 122 | 0.47 s | — | Oct 22 |
| Veniceprecision not read: host terms · 1M ctx · 128K out | $0.15 | $0.50 | $0.03 | 62 | 2.16 s | 96.1% | Oct 23 |
| Crusoefp4 (fp8) · 1M ctx | $0.15 | $0.50 | $0.03 | 79 | 0.90 s | 98.5% | Oct 23 |
| AtlasCloudfp8 · 1M ctx · 128K out | $0.15 | $0.50 | $0.03 | 28 | 6.21 s | 96.4% | Oct 23 |
| Cohereprecision not read: host terms · 1M ctx · 128K out | $0.15 | $0.50 | $0.03 | 31 | 2.51 s | 83.6% | Oct 23 |
| Cloudflareprecision not stated · 1M ctx | $0.30 | $1.00 | $0.03 | — | — | 99.3% | Oct 8 |
| DekaLLM1M ctx | $0.10 | $1.00 | $0.04 | — | — | 99.6% | Oct 11 |
| Rekaprecision not read: host terms · 1M ctx | $0.08 | $2.40 | $0.04 | 35 | 0.90 s | 99.9% | Oct 23 |
| Morphfp8 · 1M ctxreprices often | $0.16 | $0.65 | $0.045 | 13 | 2.09 s | 100% | Oct 23 |
| NextBitfp4 (fp8) · 1M ctx · 125K out | $0.15 | $0.75 | $0.05 | 44 | 1.76 s | 99.2% | Oct 23 |
| Inceptronfp8 · 1M ctxreprices often | $0.18 | $0.69 | $0.075 | 75 | 0.41 s | 99.8% | Oct 23 |
10 of 29 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 |
|---|---|---|---|---|---|---|
| StreamLake | fp8 | 1000K | 125K | yes | yes | yes |
| Pareto Inference | not stated by the host | not stated by the host | 128K | — | — | — |
| GMICloud | fp8 | 1M | not read: host terms | yes | no | yes |
| Io Net | not read: host terms | 256K | 64K | — | — | — |
| InferenceNet | fp4 | 1M | 256K | yes | yes | yes |
| Phala | nvfp4 | 1M | 128K | yes | yes | yes |
| Relace | not stated by the host | 1M | 128K | yes | no | yes |
| Novita AI | fp8 | 1M | 128K | yes | no | yes |
| DeepInfra | fp4 | 1M | not stated by the host | yes | yes | yes |
| Z.AI | fp8 | 1M | 128K | yes | no | yes |
| Modal | nvfp4 | 1M | not stated by the host | yes | yes | yes |
| Fireworks AI | not checked yet | 1M | not checked yet | yes | yes | yes |
| Friendli | not read: host terms | 1M | not read: host terms | yes | yes | yes |
| SiliconFlow | fp8 | 1M | 256K | yes | no | yes |
| DigitalOcean | not checked yet | 1M | 128K | yes | yes | yes |
| Together AI | fp4 | 1.05M | not stated by the host | yes | yes | yes |
| Parasail | fp4 | 1M | not read: host terms | yes | yes | yes |
| BaseTen | fp8 | 1M | 128K | yes | yes | yes |
| Venice | not read: host terms | 1M | 128K | yes | yes | yes |
| CoreWeave | nvfp4 | 1M | not checked yet | yes | yes | yes |
| Crusoe | fp4 | 1M | not stated by the host | yes | yes | yes |
| AtlasCloud | fp8 | 1M | 128K | yes | no | yes |
| Cohere | not read: host terms | 1M | 128K | yes | yes | yes |
| Morph | fp8 | 1M | not stated by the host | yes | yes | yes |
| Inceptron | fp8 | 1M | not stated by the host | yes | yes | yes |
| NextBit | fp4 | 1M | 125K | yes | yes | yes |
| DekaLLM | not checked yet | 1M | not checked yet | yes | yes | yes |
| Cloudflare | not stated by the host | 1M | not stated by the host | — | — | — |
| Reka | not read: host terms | 1M | not read: host terms | yes | yes | 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-5.3-Flash spec sheet, field by field, with sources
GLM-5.3-Flash 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
- 1,048,576 tokens
- Parameters
- 321,323,031,390≈ 321.3B
- Active experts
- 8 of 288 active per token
- Weight precision
- fp8
- 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 · image · video
- Max output (official)
- 128,000 tokensmaker's docs · read 2026-09-19
- Release date
- to verify≈ August 26, 2026 third-party catalogue — not the maker's figure
- Maker lifecycle status
- to verify
Also declared in the weights
- Architecture class
- Glm5NextForConditionalGeneration
- Layers
- 45
- Vocabulary
- 154,880 tokens
- Weights first published
- August 25, 2026
- Weights last updated
- September 7, 2026
- Languages declared
- English · Chinese
- Paper
- arXiv:2602.15763
zai-org/GLM-5.3-Flash · config.json @ eb9eb20 · read 2026-10-08 · 7 of 10 fields published by the maker
| Date | Provider | Input | Output | Change |
|---|---|---|---|---|
| Oct 23 | Morph | $0.16 → $0.16 | $0.65 → $0.65 | 0% |
| Oct 23 | Reka | $0.08 → $0.08 | $2.00 → $2.40 | +19% |
| Oct 22 | Inceptron | $0.10 → $0.18 | $0.69 → $0.69 | +10% |
| Oct 22 | NextBit | $0.12 → $0.15 | $0.40 → $0.75 | +73% |
| Oct 22 | Reka | $0.10 → $0.08 | $2.40 → $2.00 | −17% |
| Oct 22 | StreamLake | $0.0588 → $0.0435 | $0.20 → $0.14 | −26% |
| Oct 21 | DeepInfra | $0.075 → $0.15 | $0.25 → $0.50 | +100% |
| Oct 21 | Reka | $0.06 → $0.10 | $1.60 → $2.40 | +51% |
| Oct 21 | StreamLake | $0.0738 → $0.0588 | $0.25 → $0.20 | −20% |
| Oct 20 | GMICloud | $0.09 → $0.075 | $0.30 → $0.25 | −17% |
| Oct 20 | NextBit | $0.25 → $0.12 | $0.70 → $0.40 | −45% |
| Oct 20 | StreamLake | $0.0748 → $0.0738 | $0.25 → $0.25 | −1% |
| Oct 19 | StreamLake | $0.075 → $0.0748 | $0.25 → $0.25 | 0% |
| Oct 18 | Inceptron | $0.10 → $0.10 | $0.86 → $0.69 | −18% |
| Oct 17 | Inceptron | $0.0801 → $0.10 | $0.55 → $0.86 | +53% |
| Oct 17 | StreamLake | $0.0435 → $0.075 | $0.14 → $0.25 | +72% |
| Oct 16 | Inceptron | $0.11 → $0.0801 | $0.55 → $0.55 | −4% |
| Oct 16 | NextBit | $0.17 → $0.25 | $0.55 → $0.70 | +33% |
| Oct 14 | Inceptron | $0.0783 → $0.11 | $0.55 → $0.55 | +5% |
| Oct 13 | Inceptron | $0.096 → $0.0783 | $0.55 → $0.55 | −3% |
| Oct 13 | StreamLake | $0.0598 → $0.0435 | $0.20 → $0.14 | −27% |
| Oct 12 | Inceptron | $0.12 → $0.096 | $0.55 → $0.55 | −4% |
| Oct 12 | StreamLake | $0.069 → $0.0598 | $0.23 → $0.20 | −13% |
| Oct 11 | StreamLake | $0.0838 → $0.069 | $0.28 → $0.23 | −18% |
| Oct 10 | Inceptron | $0.23 → $0.12 | $0.60 → $0.55 | −19% |
| Oct 10 | StreamLake | $0.087 → $0.0838 | $0.29 → $0.28 | −4% |
| Oct 9 | Reka | $0.15 → $0.06 | $0.50 → $1.60 | +155% |
| Oct 8 | Pareto Inference | $0.09 → $0.06 | $0.30 → $0.20 | −33% |
| Oct 8 | Reka | $0.30 → $0.15 | $2.00 → $0.50 | −72% |
| Oct 7 | Reka | $0.40 → $0.30 | $2.00 → $2.00 | −4% |
| Oct 6 | Inceptron | $0.16 → $0.23 | $0.45 → $0.60 | +35% |
| Oct 6 | Reka | $0.15 → $0.40 | $0.50 → $2.00 | +269% |
| Oct 5 | DekaLLM | $0.10 → $0.10 | $0.40 → $1.00 | +120% |
| Oct 5 | Inceptron | $0.23 → $0.16 | $0.45 → $0.45 | −10% |
| Oct 5 | StreamLake | $0.0838 → $0.087 | $0.28 → $0.29 | +4% |
| Oct 4 | Io Net | $0.14 → $0.11 | $0.45 → $0.38 | −17% |
| Oct 3 | StreamLake | $0.087 → $0.0838 | $0.29 → $0.28 | −4% |
| Oct 2 | Io Net | $0.15 → $0.14 | $0.50 → $0.45 | −10% |
| Oct 1 | AtlasCloud | $0.12 → $0.15 | $0.39 → $0.50 | +30% |
| Oct 1 | Inceptron | $0.15 → $0.23 | $0.45 → $0.45 | +13% |
| Oct 1 | StreamLake | $0.13 → $0.087 | $0.42 → $0.29 | −32% |
| Sep 30 | Inceptron | $0.12 → $0.15 | $0.45 → $0.45 | +5% |
| Sep 30 | Io Net | $0.14 → $0.15 | $0.46 → $0.50 | +9% |
| Sep 30 | Pareto Inference | $0.03 → $0.09 | $0.10 → $0.30 | +200% |
| Sep 29 | AtlasCloud | $0.15 → $0.12 | $0.50 → $0.39 | −23% |
| Sep 29 | Inceptron | $0.11 → $0.12 | $0.45 → $0.45 | +2% |
| Sep 29 | Io Net | $0.14 → $0.14 | $0.45 → $0.46 | +2% |
| Sep 29 | StreamLake | $0.14 → $0.13 | $0.47 → $0.42 | −10% |
| Sep 28 | Io Net | $0.15 → $0.14 | $0.50 → $0.45 | −10% |
| Sep 26 | Inceptron | $0.15 → $0.11 | $0.50 → $0.45 | −14% |
| Sep 25 | Inceptron | $0.09 → $0.15 | $0.28 → $0.50 | +76% |
| Sep 25 | Io Net | $0.11 → $0.15 | $0.38 → $0.50 | +33% |
| Sep 25 | Modal | $0.45 → $0.15 | $1.50 → $0.50 | −67% |
| Sep 24 | Cloudflare | $0.15 → $0.30 | $0.50 → $1.00 | +100% |
| Sep 24 | Inceptron | $0.15 → $0.09 | $0.50 → $0.28 | −43% |
| Sep 24 | Io Net | $0.15 → $0.11 | $0.50 → $0.38 | −25% |
| Sep 24 | NextBit | $0.18 → $0.17 | $0.60 → $0.55 | −8% |
| Sep 23 | GMICloud | $0.075 → $0.09 | $0.25 → $0.30 | +20% |
| Sep 21 | GMICloud | $0.10 → $0.075 | $0.35 → $0.25 | −29% |
| Sep 18 | NextBit | $0.20 → $0.18 | $0.68 → $0.60 | −11% |
| Sep 18 | Reka | $0.23 → $0.15 | $0.75 → $0.50 | −33% |
| Sep 17 | NextBit | $0.18 → $0.20 | $0.59 → $0.68 | +14% |
| Sep 17 | Reka | $0.13 → $0.23 | $0.44 → $0.75 | +70% |
| Sep 17 | StreamLake | $0.10 → $0.14 | $0.35 → $0.47 | +34% |
| Sep 16 | GMICloud | $0.11 → $0.10 | $0.38 → $0.35 | −7% |
| Sep 16 | StreamLake | $0.11 → $0.10 | $0.37 → $0.35 | −7% |
| Sep 13 | NextBit | $0.15 → $0.18 | $0.50 → $0.59 | +18% |
| Sep 13 | Reka | $0.15 → $0.13 | $0.50 → $0.44 | −12% |
| Sep 12 | StreamLake | $0.14 → $0.11 | $0.45 → $0.37 | −17% |
| Sep 11 | Modal | $0.15 → $0.45 | $0.50 → $1.50 | +200% |
| Sep 11 | Z.AI | $0.075 → $0.15 | $0.25 → $0.50 | +100% |
| Sep 10 | GMICloud | $0.075 → $0.11 | $0.25 → $0.38 | +50% |
| Sep 9 | Novita AI | $0.075 → $0.15 | $0.25 → $0.50 | +100% |
Every published change of input or output price, newest first · change on input + output · full series: pricing-history.csv.
| Provider | Chatbot 1,000 in / 500 out | RAG / long context 8,000 in / 500 out | Batch summarize 4,000 in / 1,000 out |
|---|---|---|---|
| StreamLake | $0.12 | $0.42 | $0.32 |
| Pareto Inference | $0.16 | $0.58 | $0.44 |
| GMICloud | $0.20 | $0.72 | $0.55 |
| Io Net | $0.30 | $1.09 | $0.82 |
| InferenceNet | $0.33 | $1.03 | $0.85 |
| Phala | $0.34 | $1.23 | $0.93 |
| Relace | $0.33 | $0.85 | $0.80 |
| Novita AI | $0.40 | $1.45 | $1.10 |
| DeepInfra | $0.40 | $1.45 | $1.10 |
| Z.AI | $0.40 | $1.45 | $1.10 |
| Modal | $0.40 | $1.45 | $1.10 |
| Fireworks AI | $0.40 | $1.45 | $1.10 |
| Friendli | $0.40 | $1.45 | $1.10 |
| SiliconFlow | $0.40 | $1.45 | $1.10 |
| DigitalOcean | $0.40 | $1.45 | $1.10 |
| Together AI | $0.40 | $1.45 | $1.10 |
| Parasail | $0.40 | $1.45 | $1.10 |
| BaseTen | $0.40 | $1.45 | $1.10 |
| Venice | $0.40 | $1.45 | $1.10 |
| CoreWeave | $0.40 | $1.45 | $1.10 |
| Crusoe | $0.40 | $1.45 | $1.10 |
| AtlasCloud | $0.40 | $1.45 | $1.10 |
| Cohere | $0.40 | $1.45 | $1.10 |
| Morph | $0.49 | $1.60 | $1.29 |
| Inceptron | $0.53 | $1.78 | $1.41 |
| NextBit | $0.53 | $1.57 | $1.35 |
| DekaLLM | $0.60 | $1.30 | $1.40 |
| Cloudflare | $0.80 | $2.90 | $2.20 |
| Reka | $1.28 | $1.84 | $2.72 |
What each host bills for 1,000 requests of a representative job