RTX 5090 · 32 GB VRAM
NVIDIA · RTX 5090
158 text models run well · 1,792 GB/s bandwidth
Models on this hardware
Start here
File UD-Q4_K_M · 16.9 GB · unsloth/gemma-4-26B-A4B-it-GGUF · What is Q4?
Memory Needs ~19 GB · ~21 GB on RTX 5090 · 32 GB VRAM, including ~2 GB the system keeps
Measured 231.4 tok/s · UD-Q4_K_M · llama.cpp CUDA · source · What is tok/s?
ollama run hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-Q4_K_M- Install Ollama
- Run the command above
- Ask:
Explain what a context window is, in two sentences.
File Q4_K_M · 16.5 GB · prithivMLmods/Smaug-Mini-GGUF · What is Q4?
Memory Needs ~21 GB · ~23 GB on RTX 5090 · 32 GB VRAM, including ~2 GB the system keeps
ollama run hf.co/prithivMLmods/Smaug-Mini-GGUF:Q4_K_M- Install Ollama
- Run the command above
- Ask:
Explain what a context window is, in two sentences.
Text
Text · LiveBench · overall
0–100 · higher is better
- Color = publisher
Source: LiveBench 2026-06-25 · collected 2026-09-26 · Open models without a score are not drawn (no score ≠ 0).
| Smaug MiniQ4 · published params · source | 76.9 | Not on the LMArena text leaderboard of 2026-10-22 | ~23 GB | Q4 · published params |
|---|---|---|---|---|
| Qwen3.8 27BQ4 · published params · source | 75.3 | 1,437 ± 5 | ~23 GB | Q4 · published params |
| Qwen3.6 27BQ4 · published params · source | 64.0 | Not on the LMArena text leaderboard of 2026-10-22 | ~23 GB | Q4 · published params |
| Qwen3 14BQ4 · published params · source | No comparable published result | Not on the LMArena text leaderboard of 2026-10-22 | ~14 GB | Q4 · published params |
| Gemma 4 31BQ4 · published params · source | No comparable published result | 1,452 ± 7 | ~26 GB | Q4 · published params |
| Gemma 4 26B A4BQ4 · published params · source | No comparable published result | 1,437 ± 7 | ~21 GB | Q4 · published params |
| Qwen3 8BQ4 · published params · source | No comparable published result | Not on the LMArena text leaderboard of 2026-10-22 | ~9 GB | Q4 · published params |
| Qwen3.5 9BQ4 · published params · source | No comparable published result | Not on the LMArena text leaderboard of 2026-10-22 | ~9 GB | Q4 · published params |
LMArena ratings in this table: LMArena 2026-10-22 · collected 2026-10-25 · dataset · CC BY 4.0
Tight fit: it runs, but leaves less than 3.2 GB free beyond the 2 GB the system keeps (the memory figure already includes that reserve).
Every reviewed text model here 37
Every one that runs here, well or slowly, scored or not; 19 of 37 have a LiveBench or LMArena result. Run well first, then tight fits, then slow; within each, by Hub downloads over 30 days, read 2026-10-25, not by quality.
| Qwen3 14B | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~14 GB | 9.0 | Apache 2.0 | 10.4M |
|---|---|---|---|---|---|---|
| Gemma 4 31B | Not in LiveBench 2026-06-25 | 1,452 | ~26 GB | 18.3 | Apache 2.0 | 8.4M |
| Gemma 4 26B A4B | Not in LiveBench 2026-06-25 | 1,437 | ~21 GB | 16.9 | Apache 2.0 | 8.2M |
| Qwen3 8B | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~9 GB | 5.0 | Apache 2.0 | 8.1M |
| Qwen3.8 27B | 75.3 | 1,437 | ~23 GB | 16.5 | Apache 2.0 | 7.8M |
| Qwen3.5 9B | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~9 GB | 5.7 | Apache 2.0 | 6.1M |
| gpt-oss-20b | Not in LiveBench 2026-06-25 | 1,317 | ~18 GB | 11.6 | Apache 2.0 | 5.8M |
| Llama 3.1 8B Instruct | Not in LiveBench 2026-06-25 | 1,211 | ~8 GB | 4.9 | Llama 3.1 | 4.8M |
| Qwen3 1.7B | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~4 GB | 1.1 | Apache 2.0 | 4.1M |
| Gemma 4 E4B | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~8 GB | 5.0 | Apache 2.0 | 4M |
| Llama 3.2 1B Instruct | Not in LiveBench 2026-06-25 | 1,111 | ~3 GB | 0.8 | Llama 3.2 | 3.5M |
| Gemma 4 E2B | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~6 GB | 3.1 | Apache 2.0 | 3M |
| Qwen3 4B Instruct 2507 | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~5 GB | 2.5 | Apache 2.0 | 2.6M |
| Qwen3.6 27B | 64.0 | Not on the LMArena text leaderboard of 2026-10-22 | ~23 GB | 16.8 | Apache 2.0 | 2.4M |
| Gemma 4 12B | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~11 GB | 7.1 | Apache 2.0 | 1.7M |
| Llama 3.2 3B Instruct | Not in LiveBench 2026-06-25 | 1,167 | ~5 GB | 2.0 | Llama 3.2 | 1.1M |
| Qwen3 30B A3B Instruct 2507 | Not in LiveBench 2026-06-25 | 1,382 | ~25 GB | 18.6 | Apache 2.0 | 1.1M |
| GLM-4.7 Flash | Not in LiveBench 2026-06-25 | 1,365 | ~25 GB | 18.3 | MIT | 953.7K |
| Nemotron 3.5 Lightning 30B A3B | Not in LiveBench 2026-06-25 | Arena rated the NVFP4 build, NVIDIA-Nemotron-3.5-Lightning-30B-A3B (NVFP4); this entry is the BF16 reference, which NVIDIA reports with different scores | ~25 GB | 18.9 | OpenMDW 1.1 | 861.5K |
| Qwen3 Coder 30B A3B | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~25 GB | 18.6 | Apache 2.0 | 745K |
| SmolLM3 3B | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~5 GB | 1.9 | Apache 2.0 | 718.7K |
| Phi-4 | Not in LiveBench 2026-06-25 | 1,256 | ~13 GB | 8.0 | MIT | 368.3K |
| DeepSeek R1 0528 Qwen3 8B | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~9 GB | 5.0 | MIT | 355.8K |
| Phi-4 mini | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~5 GB | 2.5 | MIT | 342.8K |
| Olmo 3 7B Instruct | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~8 GB | 4.5 | Apache 2.0 | 262.5K |
| Devstral Small 2 24B | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~20 GB | 14.3 | Apache 2.0 | 195.6K |
| Mistral Small 3.2 24B | Not in LiveBench 2026-06-25 | 1,356 | ~21 GB | 14.3 | Apache 2.0 | 181.4K |
| Ministral 3 14B | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~13 GB | 8.2 | Apache 2.0 | 126.5K |
| Ministral 3 3B | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~5 GB | 2.1 | Apache 2.0 | 87.4K |
| Ministral 3 8B | Not in LiveBench 2026-06-25 | Not on the LMArena text leaderboard of 2026-10-22 | ~9 GB | 5.2 | Apache 2.0 | 79.4K |
| Granite 4.2 8B | Not in LiveBench 2026-06-25 | 1,290 | ~8 GB | 5.3 | Apache 2.0 | 78.3K |
| Granite 4.2 3B | Not in LiveBench 2026-06-25 | 1,290 | ~5 GB | 2.2 | Apache 2.0 | 39.8K |
| Granite 4.2 30B | Not in LiveBench 2026-06-25 | 1,340 | ~25 GB | 17.7 | Apache 2.0 | 39.3K |
| Olmo 3 32B Think | Not in LiveBench 2026-06-25 | 1,306 | ~26 GB | 19.5 | Apache 2.0 | 13.1K |
| Smaug Mini | 76.9 | Not on the LMArena text leaderboard of 2026-10-22 | ~23 GB | 16.5 | Apache 2.0 | 817 |
| Qwen3.5 35B A3B | Not in LiveBench 2026-06-25 | 1,394 | ~29 GBTight fit | 22.0 | Apache 2.0 | 1.5M |
| Nemotron 3 Nano 30B A3B | Not in LiveBench 2026-06-25 | 1,314 | ~32 GBTight fit | 24.6 | NVIDIA Nemotron Open Model | 1.4M |
Also runs slowly
Above the 32 GB of VRAM, the layers that do not fit run on the CPU from PC RAM, up to 56 GB in all. 32 GB of PC RAM assumed: this configuration lists none. Planning estimate, not a speed figure.
- Smaug-Mixtral-v0.1abacusai~38 GBQ4 · tensor count
- StableBeluga1-Deltastabilityai~51 GBQ4 · tensor count
- Llama-2-70b-hfmeta-llama~54 GBQ4 · tensor count
- Llama-2-70b-chat-hfmeta-llama~54 GBQ4 · tensor count
- japanese-stablelm-instruct-beta-70bstabilityai~54 GBQ4 · tensor count
- japanese-stablelm-base-beta-70bstabilityai~54 GBQ4 · tensor count
- Giraffe-v2-70b-32kabacusai~54 GBQ4 · tensor count
- CodeLlama-70b-Instruct-hfmeta-llama~54 GBQ4 · tensor count
- CodeLlama-70b-hfmeta-llama~54 GBQ4 · tensor count
- CodeLlama-70b-Python-hfmeta-llama~54 GBQ4 · tensor count
- Llama-3.3-70B-Instructmeta-llama~55 GBQ4 · tensor count
- Llama-3.1-70B-Instructmeta-llama~55 GBQ4 · tensor count
- Meta-Llama-3-70Bmeta-llama~55 GBQ4 · tensor count
- Llama-3.1-70Bmeta-llama~55 GBQ4 · tensor count
- DeepSeek-R1-Distill-Llama-70Bdeepseek-ai~55 GBQ4 · tensor count
- Meta-Llama-3-70B-Instructmeta-llama~55 GBQ4 · tensor count
- Smaug-Llama-3-70B-Instructabacusai~55 GBQ4 · tensor count
- Smaug-Llama-3-70B-Instruct-32Kabacusai~55 GBQ4 · tensor count
- Infinity-Instruct-7M-Gen-Llama3_1-70BBAAI~55 GBQ4 · tensor count
- Infinity-Instruct-3M-0625-Llama3-70BBAAI~55 GBQ4 · tensor count
- Infinity-Instruct-3M-0613-Llama3-70BBAAI~55 GBQ4 · tensor count
- Dracarys-Llama-3.1-70B-Instructabacusai~55 GBQ4 · tensor count
- Llama-3-Giraffe-70Babacusai~55 GBQ4 · tensor count
- Dracarys2-Llama-3.1-70B-Instructabacusai~55 GBQ4 · tensor count
- Llama-3-Giraffe-70B-Instructabacusai~55 GBQ4 · tensor count
Adjust the estimate: quantization, context, reserve
Vision
Vision · LMArena
rating · higher is better
- Color = publisher
Source: LMArena 2026-10-19 · collected 2026-10-25 · dataset · CC BY 4.0 · Models without a rating are left out, not scored as zero.
| Gemma 4 31BQ4 · published params · source | 1,262 ± 6 | ~26 GB | Q4 · published params |
|---|---|---|---|
| Qwen3.8 27BQ4 · published params · source | 1,242 ± 8 | ~23 GB | Q4 · published params |
| Gemma 4 26B A4BQ4 · published params · source | 1,242 ± 6 | ~21 GB | Q4 · published params |
| gemma-3-27b-itQ4 · Hugging Face count · source | 1,160 ± 8 | ~23 GB | Q4 · Hugging Face count |
| Qwen2.5-VL-32B-InstructQ4 · Hugging Face count · source | 1,119 ± 15 | ~28 GB | Q4 · Hugging Face count |
| Molmo2-8BQ4 · Hugging Face count · source | 1,112 ± 20 | ~9 GB | Q4 · Hugging Face count |
| Qwen2-VL-7B-InstructQ4 · Hugging Face count · source | 1,031 ± 11 | ~9 GB | Q4 · Hugging Face count |
| Molmo-7B-D-0924Q4 · Hugging Face count · source | 994 ± 14 | ~8 GB | Q4 · Hugging Face count |
LMArena ratings in this table: LMArena 2026-10-19 · collected 2026-10-25 · dataset · CC BY 4.0
Tight fit: it runs, but leaves less than 3.2 GB free beyond the 2 GB the system keeps (the memory figure already includes that reserve).
12 more vision models in the catalog have no usable memory figure or parameter count for this profile. Browse Vision models
Image
Image · Prompt alignment · GenEval
0–1 · higher is better
Source: GenEval · 2023 · reviewed 2026-10-05 · Published quality in the original benchmark preset; not measured on this hardware.
| Stable Diffusion XL 1.0Platform path documented · stable-diffusion.cpp · source | 0.55 | Not on the LMArena text-to-image leaderboard of 2026-10-07 | Documented path without a memory figure | Platform path documented |
|---|---|---|---|---|
| Stable Diffusion 1.5Platform path documented · stable-diffusion.cpp · source | 0.43 | Not on the LMArena text-to-image leaderboard of 2026-10-07 | Documented path without a memory figure | Platform path documented |
| BagelPublisher minimum · CUDA · source | No comparable published result | 898 ± 6 | ~32 GB | Publisher minimum |
| Step1X-Edit-v1p216-bit est. · params · source | No comparable published result | Not reviewed for LMArena | ~32 GB | 16-bit est. · params |
| FLUX.1 Krea [dev]16-bit est. · params · source | No comparable published result | Not on the LMArena text-to-image leaderboard of 2026-10-07 | ~31 GB | 16-bit est. · params |
| FLUX.1-Kontext-dev16-bit est. · params · source | No comparable published result | Not reviewed for LMArena | ~31 GB | 16-bit est. · params |
| SRPO16-bit est. · params · source | No comparable published result | Not on the LMArena text-to-image leaderboard of 2026-10-07 | ~31 GB | 16-bit est. · params |
| FLUX.2 [klein] 9BPublisher minimum · CUDA · source | No comparable published result | 1,070 ± 3 | ~29 GB | Publisher minimum |
LMArena ratings in this table: LMArena 2026-10-07 · collected 2026-10-25 · dataset · CC BY 4.0
17 more image models in the catalog have no usable memory figure or parameter count for this profile. Browse Image models
Video
| LTX-2.5 FastPublisher minimum · CUDA · source | ~32 GB | Publisher minimum |
|---|---|---|
| LTX-2.5 ProPublisher minimum · CUDA · source | ~32 GB | Publisher minimum |
| Open-Sora-v216-bit est. · params · source | ~29 GB | 16-bit est. · params |
| mochi-1-preview16-bit est. · params · source | ~26 GB | 16-bit est. · params |
| Wan2.2-TI2V-5BConditional path · CUDA · source | ~24 GB | Conditional path |
| Wan2.2-TI2V-5B-DiffusersConditional path · CUDA · source | ~24 GB | Conditional path |
| HunyuanVideo-1.516-bit est. · params · source | ~22 GB | 16-bit est. · params |
| CogVideoX-5b-I2V16-bit est. · params · source | ~16 GB | 16-bit est. · params |
How this list works Documented execution paths on this hardware first, then publisher minimums (CUDA or Apple MPS).
20 more video models in the catalog have no usable memory figure or parameter count for this profile. Browse Video models
Audio
Audio · Word error rate · Open ASR
% · lower is better
Source: Word error rate · Open ASR · reviewed 2026-10-20 · Published quality in the original benchmark preset; not measured on this hardware.
| Whisper large-v3-turboPlatform path documented · whisper.cpp / CUDA · source | 7.83 | Documented path without a memory figure | Platform path documented |
|---|---|---|---|
| Whisper large-v3Platform path documented · whisper.cpp / CUDA · source | 7.44 | Documented path without a memory figure | Platform path documented |
| Kimi-Audio-7B16-bit est. · params · source | No comparable published result | ~26 GB | 16-bit est. · params |
| Kimi-Audio-7B-Instruct16-bit est. · params · source | No comparable published result | ~26 GB | 16-bit est. · params |
| AuKPublisher minimum · CUDA · source | No comparable published result | ~25 GB | Publisher minimum |
| AuK-FlashPublisher minimum · CUDA · source | No comparable published result | ~25 GB | Publisher minimum |
| Fish Audio S2 ProPublisher minimum · CUDA · source | No comparable published result | ~24 GB | Publisher minimum |
| MiniMax-Music3Publisher minimum · CUDA · source | No comparable published result | ~24 GB | Publisher minimum |
33 more audio models in the catalog have no usable memory figure or parameter count for this profile. Browse Audio models
Embeddings
Embeddings · MTEB (Multilingual, v2) · mean(task)
0–100 · higher is better
- Color = publisher
| Qwen3-Embedding 8B8-bit · published params · source | 70.6 | ~12 GB | 8-bit · published params |
|---|---|---|---|
| Qwen3-Embedding 4B8-bit · published params · source | 69.5 | ~7 GB | 8-bit · published params |
| Harrier OSS v1 (0.6B)8-bit · published params · source | 69.0 | ~3 GB | 8-bit · published params |
| Qwen3-Embedding 0.6B8-bit · published params · source | 64.3 | ~3 GB | 8-bit · published params |
| Multilingual E5 Large Instruct8-bit · published params · source | 63.2 | ~3 GB | 8-bit · published params |
| EmbeddingGemma 28-bit · published params · source | 61.4 | ~3 GB | 8-bit · published params |
| EmbeddingGemma (300M)8-bit · published params · source | 61.1 | ~3 GB | 8-bit · published params |
| BGE-M38-bit · published params · source | 59.6 | ~3 GB | 8-bit · published params |
Runs well within 32 GB of VRAM · planning estimates, not measured speed
Measurements
Published tokens per second, exactly as measured; never estimated.
Specifications
12 sourced fields; the source is named once per group, and a field from another source links to it. Official specifications · Memory specification checked: 2026-09-30 · manufacturer source · Sourced facts JSON
Marker: the best LiveBench-scored text model that runs well here. No other text model would run well with the next memory size. Q4 planning estimate, not a measured fit.
Selected configuration
Latest check · Manufacturer source
- Memory capacity
- 32 GB
- Memory type
- 32 GB GDDR7
- Chip / processor
- Blackwell · RTX 5090
- CUDA cores
- 21,760
- GPU power (reference specification)
- 575 W
- Memory interface
- 512 bit
- CUDA compute capability
- 12.0
Family specification
Latest check · Manufacturer source
- Display connectors (reference design; OEM may vary)
- 3 × DisplayPort 2.1b · HDMI 2.1b
- Card length
- 304 mm
One CUDA GPU with 32 GB dedicated VRAM. A model below 32 GB in the generic Q4 estimate is a planning candidate, not a tested fit. CPU offload needs system RAM and may reduce speed.
Photo: Official NVIDIA reference image; partner board designs vary. · Manufacturer source
Memory bandwidth is not measured model speed; verify runtime and checkpoint compatibility. Hardware specification reviewed: 2026-10-13.
How estimates work
- Candidates
- Reviewed pages use published parameter counts; automatic text candidates use official repository tensor counts (packed quantized weights and unsupported architectures excluded).
- Planning
- Q4 weights + 20% allowance + reserve: 25% of unified or system memory (at least 2 GB, at most 8 GB; 2 GB here) or 2 GB per dedicated GPU. Adjustable in “Adjust the estimate”. No KV cache is calculated from context.
- Fit tiers
- Runs well: the estimate fits the fast memory (VRAM, or unified memory within the macOS GPU share). Runs slowly: within the machine's memory but outside the fast part; a graphics card offloads layers to PC RAM (the listed RAM, or 32 GB assumed). Doesn't run: above both. Tiers are planning estimates, never a speed.
- Vision
- Vision-language models use the Q4 planning estimate from the parameter count (published by the model page, or read from Hugging Face). Classic vision models (classify, detect, segment) use a 16-bit estimate. No vision benchmark is tracked, so rows carry no score.
- Media and embeddings
- Media rows use documented execution paths (exact hardware or platform) and publisher minimums; a published requirement above this memory moves a model to the next tier. Without either, a 16-bit estimate from the published parameter count (× 1.2 + 2 GB), labelled as such. Embeddings use 8-bit weights + 20% + 2 GB.
Local AI basics: parameters, Q4 and memory · Full model catalog · Data and methodology
Alternatives
Side-by-side comparisons, then other machines on the same memory ladder (VRAM); each page lists the models that fit it.
