GLM-5.3-Flash 320B
GLM-5.3-Flash 320B needs roughly 218.5 GB VRAM at Q4_K_M quantization (717.0 GB at FP16). 8 GPUs we track can run it fully in VRAM at 8k context.
8 GPUs run this natively · 2 with CPU offload
- Q4_K_M total
- 218.5 GB
- at 8k context
- Smallest GPU
- 256 GB
- Apple M5 Ultra (256GB), at Q4_K_M
- KV cache, full context
- 23.6 GB
- 11 of 45 layers cache
- Inputs
- Text · Image · Video
- MIT, released August 2026
- Active experts
- 8 routed + 1 shared
- of 288 total
- Active parameters
- 18B
- 5.6% of the 320B total
- Context ceiling
- 1,048,576 tokens
- native, no scaling needed
GLM-5.3-Flash 320B is a Mixture of Experts (MoE) model with 320B total parameters but only 18B active per token developed by Z.ai. Released 25 August 2026 under MIT, the first natively multimodal model in the GLM-5 series (text, image, and video input) and the first GLM release with hybrid attention: 45 layers, eleven repeats of three linear-attention layers followed by one DeepSeek Sparse Attention layer (Multi-head Latent Attention, a 512-dimension compressed cache), plus one closing linear-attention layer. It's a mixture of experts, 288 routed plus 1 shared expert with 8 routed active per token (18B of 320B total active), trained on a 30-trillion-token multimodal corpus and using Manifold-Constrained Hyper-Connections (mHC) for scaling stability. Native context is 1,048,576 tokens.
To run GLM-5.3-Flash 320B locally: a Q4_K_M GGUF is roughly 194.9GB of weights, about 218.5GB in total at 8k context, already past a single 180GB B200's usable capacity. The smallest single machine this site tracks that actually fits it is a 256GB Apple M3 Ultra (a 288GB NVIDIA B300 clears it too). The hybrid attention is what keeps that same 256GB Mac usable at the model's full 1,048,576-token native context too: the real 11-of-45 caching layers cost 23.6GB of KV cache at that length, versus 96.6GB if every layer cached conventionally, and the full-context total at Q4_K_M comes to 244.7GB, still inside the 256GB Mac. An all-attention version of the same model would need 326.5GB at that context, past both the 256GB Mac and the 288GB B300, reachable only on a 512GB machine. As a MoE model, inference speed depends on active parameters (18B) rather than total size.
GPQA Diamond 91.2%, independently measured by Artificial Analysis (Z.ai's own release chart doesn't report GPQA at all). On the benchmarks Z.ai does report, its own chart puts GLM-5.3-Flash ahead of GLM-5.2 on every one shown: Terminal-Bench 2.1 84.3 (vs 81.0), DeepSWE v1.1 63.4 (vs 46.2), AutomationBench 48.8 (vs 26.2), Agents' Last Exam 26.3 (vs 20.4), HLE with tools 55.3 (vs 54.7), and GDPVal-AA v2 1773 (vs 1504, also scored by Artificial Analysis). The same chart places it close to, and on several benchmarks past, Claude Opus 4.8, GPT-5.6 Terra, and Gemini 3.7 Flash, at roughly a tenth of Opus-class API pricing. Z.ai's own figures are vendor-reported against its own chosen comparison set; no independent lab had re-run them at time of writing.
Eleven sparse-attention layers carry the whole KV cache
GLM-5.3-Flash is a mixture of experts: 288 routed experts exist in the weights, and a router picks 8 plus one shared expert for each token, so 18 of the 320 billion parameters actually compute per step. Its 45 layers are also the first hybrid-attention stack in the GLM-5 line: a block of three linear-attention layers (gated recurrent, labelled "KDA" in the model's own config) followed by one DeepSeek Sparse Attention layer, repeated eleven times, with one more linear-attention layer closing out the stack.
11 of 45 layers keep a KV cache that grows with the full context. The other 34 hold a fixed-size recurrent state, so their memory cost does not move as the context window grows.
The 34 linear-attention layers carry a fixed-size recurrent state that costs the same at 2,000 tokens or 1,048,576. Only the 11 DeepSeek Sparse Attention layers store a per-token cache, and even that cache is compressed: Multi-head Latent Attention keeps one 512-dimension latent per token instead of separate per-head keys and values, the same compression approach DeepSeek's V4 line uses. A lightweight indexer then selects which of those cached entries each query actually attends to, which cuts compute, not what has to stay resident in memory.
What the full 1,048,576-token window actually costs
GLM-5.2 reached 1M context with its own sparse-attention design (IndexShare) but kept every layer caching. GLM-5.3-Flash goes further: only a quarter of its layers cache at all.
KV cache only, at FP16. Weights and activation overhead sit on top of these figures.
Filling the full 1,048,576-token window costs 23.6 GB of KV cache with the real hybrid stack, versus 96.6 GB if every layer cached, a roughly 4x reduction. Add Q4_K_M's 194.9 GB of weights and the full-context total is 244.7 GB, which fits a 256 GB Apple M3 Ultra or a 288 GB NVIDIA B300. The all-attention version of the same model would need 326.5 GB at that context, past both, reachable only on a 512 GB machine.
VRAM at each quantization
Figures below assume 8k context. This model's hybrid attention stack caches only some layers, so KV cache grows much slower than linearly as context increases.
| Quant | Weights | KV cache | Total |
|---|---|---|---|
| FP32 | 1280.0 GB | 0.18 GB | 1433.8 GB |
| BF16 | 640.0 GB | 0.18 GB | 717.0 GB |
| FP16 | 640.0 GB | 0.18 GB | 717.0 GB |
| Q8_0 | 340.2 GB | 0.18 GB | 381.2 GB |
| Q6_K | 262.7 GB | 0.18 GB | 294.4 GB |
| Q5_K_M | 227.8 GB | 0.18 GB | 255.4 GB |
| Q4_K_Mrec | 194.9 GB | 0.18 GB | 218.5 GB |
| Q3_K_M | 153.9 GB | 0.18 GB | 172.6 GB |
| Q2_K | 121.9 GB | 0.18 GB | 136.8 GB |
| NVFP4cuda | 160.0 GB | 0.18 GB | 179.4 GB |
KV cache is calculated at 8k context (FP16). Note that NVFP4 only runs on CUDA GPUs. Turn on TurboQuant in the calculator above for lower KV cache estimates.
Benchmarks
Where GLM-5.3-Flash actually pulls ahead of GLM-5.2
GPQA Diamond and Terminal-Bench 2.1 are the only benchmarks both releases share in this site's own Benchmarks fields, and both land within a point of each other: GPQA 91.2 on both (GLM-5.3-Flash's figure independently measured by Artificial Analysis; GLM-5.2's is Z.ai's own reported score), Terminal-Bench 2.1 84.3 vs 81.0. Z.ai's own release chart covers several more agentic suites with no field in this site's schema; the two below are the largest point gains on that chart.
Z.ai, GLM-5.3-Flash model card benchmark chart (huggingface.co/zai-org/GLM-5.3-Flash, 25 August 2026). Both scores are Z.ai's own vendor-reported figures; independent reproduction was not yet available at time of writing.
AutomationBench nearly doubles, 26.2 to 48.8, and DeepSWE gains 17.2 points, 46.2 to 63.4, the two largest jumps on Z.ai's own chart. Terminal-Bench 2.1 moves by a smaller 3.3 points (81.0 to 84.3), and GPQA Diamond barely moves at all, 91.2 on both releases, so the newly trained base model and hybrid attention design look like they bought the most ground on long-horizon agentic tasks specifically, not on general reasoning or coding ability.
GPUs that run GLM-5.3-Flash 320B natively (8)
- NVIDIA B300 288GBNVFP4 · 117.3 t/s
- NVIDIA B200 180GBQ2_K · 135.2 t/s
- AMD Instinct MI300XQ3_K_M · 74.6 t/s
- Apple M5 Ultra (512GB)Q8_0 · 33.3 t/s
- Apple M5 Ultra (256GB)Q4_K_M · 46.5 t/s
Show 3 more
- Apple M3 Ultra (512GB)Q8_0 · 25.4 t/s
- Apple M3 Ultra (256GB)Q4_K_M · 37.2 t/s
- Apple M2 Ultra (192GB)Q3_K_M · 42.2 t/s
Plus 2 GPUs that run it with CPU offload (slower)
- NVIDIA H200 141GBQ2_K · 124 t/s
- Intel Data Center GPU Max 1550Q2_K · 34.9 t/s
Cloud GPU Rental
Cheapest way to run GLM-5.3-Flash 320B: 4× A100 80GB at Q5_K_M, ~59 tok/s, from $4.76/hr.
Pay by the hour · no contract · pods start in about a minute.
Rent this setup on RunPod ↗ (+$5 signup credit)Prices as of Oct 13, 2026. Rates change; check RunPod for current pricing.
Affiliate link: CanItRun may earn a commission. Doesn't affect the fit calculation above.
Notes
The first natively multimodal model in the GLM-5 series and the first GLM release with hybrid attention: a newly trained base model (not a post-train of GLM-5.2, unlike the text-only GLM-5.3 flagship released the same week) combining linear-attention and DeepSeek-style sparse-attention layers, plus Manifold-Constrained Hyper-Connections (mHC) for scaling stability, trained on a 30-trillion-token multimodal corpus. It's a mixture of experts: 288 routed experts plus 1 shared, 8 routed active per token (18B of 320B active), with the first 3 of 45 layers using a dense (non-routed) feed-forward network. On Z.ai's own release chart it beats GLM-5.2 on every benchmark shown: Terminal-Bench 2.1 84.3 vs 81.0, DeepSWE v1.1 63.4 vs 46.2, AutomationBench 48.8 vs 26.2, Agents' Last Exam 26.3 vs 20.4, HLE with tools 55.3 vs 54.7, and GDPVal-AA v2 1773 vs 1504 (the last scored by Artificial Analysis). Z.ai's chart also places it near or ahead of Claude Opus 4.8, GPT-5.6 Terra, and Gemini 3.7 Flash on most of those same benchmarks, at roughly a tenth of Opus-class API pricing ($0.113 per million input tokens, $0.394 per million output). These are all Z.ai's own reported figures against its own chosen comparison set; no independent lab had re-run them at time of writing. Artificial Analysis separately measured GPQA Diamond at 91.2%, a figure Z.ai itself never published.
Compare GLM-5.3-Flash 320B with other models
So, should you actually run this?
Not on a single consumer GPU: at Q4_K_M, the quantization this site recommends for it, the weights alone are 194.9 GB, and a short 8k-token session already totals 218.5 GB, more than a single 180 GB B200 holds. What the hybrid attention buys is context budget, not a smaller weights floor: the same 256 GB unified-memory Mac (or a 288 GB B300) that holds it at 8k context still holds it at its full native 1,048,576-token window, something an all-attention design at the same weight class would make much harder. If you need a self-hosted, Opus-class coding and agentic model with a million-token window and already have that class of hardware, this is a genuinely new option; everyone else gets the same weights through Z.ai's API at roughly a tenth of Opus's price.
Frequently asked questions
- What are the VRAM requirements for GLM-5.3-Flash 320B?
- GLM-5.3-Flash 320B requires approximately 218.5 GB of VRAM at Q4_K_M quantization, 381.2 GB at Q8, and 717.0 GB at FP16. These numbers assume 8k context window; its hybrid attention stack caches far fewer than all layers, so VRAM grows much slower than linearly with context.
- How many parameters does GLM-5.3-Flash 320B have?
- GLM-5.3-Flash 320B has 320 billion total parameters, but only 18 billion are active per token thanks to its Mixture of Experts (MoE) architecture. This makes inference significantly faster than the total parameter count suggests.
- Can GLM-5.3-Flash 320B run on a 16 GB GPU?
- No. At Q4_K_M, GLM-5.3-Flash 320B needs 218.5 GB of VRAM, more than 16 GB. You will need a multi-GPU server.
- Can GLM-5.3-Flash 320B run on a 24 GB GPU?
- No. Even at Q4_K_M, GLM-5.3-Flash 320B needs 218.5 GB. Consider a multi-GPU server with 219 GB+ of combined VRAM.
- What is the smallest quantization for GLM-5.3-Flash 320B that fits in 24 GB of VRAM?
- GLM-5.3-Flash 320B cannot fit in 24 GB of VRAM at any standard quantization level. The minimum needed is 136.8 GB at Q2_K.
- What GPU do I need to run GLM-5.3-Flash 320B locally?
- You need a multi-GPU server. At Q4_K_M, GLM-5.3-Flash 320B needs 218.5 GB VRAM, more than any single consumer GPU. That's roughly 3x 80 GB datacenter GPUs (H100, A100, or similar) pooled together.
- Can I run GLM-5.3-Flash 320B in the cloud?
- Yes. The cheapest RunPod setup we found that holds GLM-5.3-Flash 320B entirely in GPU memory at 8k context is 4× A100 80GB at Q5_K_M, at roughly 59.1 tokens per second and from $4.76 per hour for the whole pod. Prices as of Oct 13, 2026; RunPod rates change, so check their pricing page for the current figure.
- Can GLM-5.3-Flash run on a single consumer GPU?
- No. Even at Q4_K_M, the quantization this site recommends for it, the weights alone are about 194.9 GB, before any KV cache or activation overhead. The smallest single machine this site tracks that actually fits it at a short 8k-token context is a 256 GB Apple M3 Ultra (a 288 GB NVIDIA B300 clears it too); even a 180 GB NVIDIA B200 falls short. Nothing in the 16-48 GB range consumer cards occupy gets remotely close.
- Is GLM-5.3-Flash the same model as GLM-5.3?
- No, despite the shared version number. GLM-5.3, released earlier the same month, is a text-only post-training refresh of the existing GLM-5.2 base model with no architecture change. GLM-5.3-Flash is a separate, newly trained model: a smaller 320B/18B-active checkpoint with a new hybrid linear/sparse-attention design and native image and video input, neither of which GLM-5.3 has.
- Why does only a quarter of the model's layers keep a KV cache?
- GLM-5.3-Flash's 45 layers alternate three linear-attention layers for every one DeepSeek Sparse Attention layer, eleven times over, plus one closing linear-attention layer. Linear attention carries a fixed-size recurrent state that doesn't grow with context, so only the 11 sparse-attention layers store a per-token cache. That's what keeps the full 1,048,576-token context to 23.6 GB of KV cache instead of the 96.6 GB an all-attention version of the same model would need.