Gemma 4 12B
Local OKExcellentA · 44
Best for: Chat, Coding, Multimodal·Pop: 80/100
Perf: est. 90 tok/s, typical error 15% · first token ~0.3s
Fits in 16 GB VRAM with room to spare. Best for chat, coding, multimodal on RTX 5070 Ti.
Best Local AI Models for RTX 5070 Ti (16GB)
The RTX 5070 Ti pairs 16GB of GDDR7 with 896 GB/s of bandwidth, the highest of any 16GB card ModelFit tracks. It matches the RTX 3090's speed on 8B models (est. 125 tok/s, typical error 5% against est. 117 tok/s) and runs 14B parameter models at est. 78 tok/s, typical error 15%. The 2026 memory shortage has lifted it above MSRP like every other tier.
Quick answer
The best local LLM for the RTX 5070 Ti is Gemma 4 12B (est. 90 tok/s, typical error 15%) on its 16GB VRAM. It uses ~8GB of VRAM; the RTX 5070 Ti handles up to 14B parameter models at Q4. A 14B model at Q4 runs at est. 78 tok/s, typical error 15%. (ModelFit, 2026)
Sizing rule: a Q4 model needs about 0.6 GB of VRAM per billion parameters, and ModelFit budgets 90% of this card's 16GB for weights, context, and KV-cache. The per-size table below uses that same budget. Other strong fits: Qwen3.5 9B Instruct (Q8) (9B, ~10.7GB) and Mellum2.1 12B-A2.5B Thinking (12.1B, ~8.1GB). On this card Gemma 4 12B runs at est. 90 tok/s, typical error 15%. What it will not run: a 32B model does not fit in VRAM alone, so it spills to system RAM over PCIe and slows sharply.
$ollama run gemma4:12b
- Top pick
- Gemma 4 12B
- Decode speed
- est. 90 tok/s, typical error 15%
- VRAM needed
- ~8 GB
Speeds are ModelFit estimates from memory bandwidth and model size, not measured benchmarks. Typical error: the median gap between the ModelFit speed model and measured runs on hardware the model was not fitted on, for this backend and model size.
Cite this page: ModelFit, RTX 5070 Ti 16GB for Local LLMs: Runs 14B Q4 (est. 78 tok/s), https://modelfit.io/gpu/rtx-5070-ti/, updated October 2026, CC BY 4.0.
Specs
VRAM
16 GB GDDR7
Speed (8B Q4)
est. 125 tok/stypical error 5%
Bandwidth
896 GB/s
Architecture
Blackwell
Max model size
Up to 14B parameter models
Compatibility
10 excellent, 0 workable
Table 1
RTX 5070 Ti estimated tokens/sec by model size
Q4_K_M · ModelFit estimate| Model size | Est. speed | Fit on 16GB |
|---|---|---|
| 7B | est. 141 tok/stypical error 5% | Fits in VRAM |
| 14B | est. 78 tok/stypical error 15% | Fits in VRAM |
| 20B MoE (3.6B active) | est. 162 tok/stypical error 15% | Fits in VRAM |
| 32B | est. 7 tok/stypical error 5% | CPU offload (slow) |
| 35B MoE (3B active) | est. 65 tok/stypical error 15% | CPU offload (slow) |
| 70B | est. 2 tok/stypical error 10% | CPU offload (slow) |
| 120B MoE (5.1B active) | est. 43 tok/stypical error 15% | CPU offload (slow) |
Fig. 1RTX 5070 Ti: estimated speed by model size
Decode speed, generic Q4_K_M model of each size at 4K context, 896 GB/s, typical error 5-15%
Fits in 16 GBCPU offload (slow)
7B141
14B78
20B MoE (3.6B active)162
32B7
35B MoE (3B active)65
70B2
120B MoE (5.1B active)43
tok/s est.ModelFit estimates, not measured benchmarks: decode speed of a generic Q4_K_M model of each size at 4K context on the RTX 5070 Ti's 896 GB/s bandwidth. MoE rows scale by active parameters (decode reads only the active experts), so a 35B MoE runs far faster than a dense 32B. "CPU offload" sizes exceed the 16GB VRAM; dense models slow to a crawl there, MoE models degrade less because hot experts stay GPU-resident. What typical error means.
Fig. 2
Context costs VRAM too.
Gemma 4 12B loads ~8 GB of weights; at 16k context the KV cache adds ~3.0 GB (still fits the ~14 GB usable VRAM), and at 64k it adds ~12.0 GB (exceeds the budget, use a smaller quant or a q8_0 KV cache).
Weights 8.0 GBKV cache · 16k context 3.0 GBHeadroom 3.4 GBRuntime reserve 1.6 GB
The model leaves about 3.4 GB of the usable VRAM budget free after weights and 16k context.
KV-cache figures assume an fp16 cache, the llama.cpp/Ollama default. Standard GQA models use a size-class estimate (8 KV heads x 128 head dim class); hybrid linear-attention models (Qwen3.5/3.6, Qwen3-Next) use the exact per-token cost from their published config, since only their sparse full-attention layers cache KV. A q8_0 KV cache roughly halves either figure. Estimates, not measurements.
Get told when a model beats Gemma 4 12B on your RTX 5070 Ti
One email when a new open-weight model is a better fit for 16 GB of VRAM, plus the Thursday weekly.
You'll also get the weekly ModelFit email. By signing up you agree to our Privacy Policy. Unsubscribe from either email anytime, in one click.
Where to buy the RTX 5070 Ti
Storage & accessories for your model library
- Internal NVMe SSD · 2TBA Gen4 M.2 drive keeps your whole GGUF and quant collection on fast local storage, loading models straight off NVMe.Check price on Amazon
- USB4 NVMe Enclosure40Gbps external storage fast enough to run models from. Pair it with an M.2 drive for a portable model vault.Check price on Amazon
ModelFit may earn a commission on purchases through these links, at no extra cost to you.
RTX 5070 Ti VRAM for AI: what actually fits?
16GB GDDR7 at 896 GB/s is the highest bandwidth of any 16GB card ModelFit tracks. The RTX 5070 Ti loads 14B models at Q4 with 5-6GB to spare and runs them at est. 78 tok/s, typical error 15%. On 8B models its bandwidth pushes tokens out at est. 125 tok/s, typical error 5%, matching the RTX 3090 (24GB, est. 117 tok/s). The 896 GB/s bandwidth is 33% higher than the RTX 5080 on a per-GB basis, which keeps token generation fast for models that fit its 16GB capacity.
Table 2
RTX 5070 Ti vs similar GPUs
| Hardware | Memory | Speed | Bandwidth |
|---|---|---|---|
| RTX 5070 | 12 GB | est. 98 tok/stypical error 5% | 672 GB/s |
| RTX 4070 Ti SUPER | 16 GB | est. 94 tok/stypical error 5% | 672 GB/s |
| RTX 5070 Ti | 16 GB | est. 125 tok/stypical error 5% | 896 GB/s |
| RTX 5080 | 16 GB | est. 132 tok/stypical error 5% | 960 GB/s |
Recommended models
registry-verified10 modelsQwen3.5 9B Instruct (Q8)
Local OKOKB · 26
Best for: Quality, Coding, Reasoning·Pop: 86/100
Perf: est. 86 tok/s, typical error 5% · first token ~1.2s
Fits in 16 GB VRAM with room to spare. Best for quality, coding, reasoning on RTX 5070 Ti.
Mellum2.1 12B-A2.5B Thinking
Local OKExcellentA · 44
Best for: Agentic coding, Small machines·Pop: 76/100
Perf: est. 224 tok/s, typical error 15% · first token ~0.1s
Fits in 16 GB VRAM with room to spare. Best for agentic coding, small machines on RTX 5070 Ti.
Gemma 4 12B (Q8)
Local OKOKB · 11
Best for: Chat, Coding, Multimodal·Pop: 80/100
Perf: est. 57 tok/s, typical error 15% · first token ~1.6s
Fits in 16 GB VRAM with room to spare. Best for chat, coding, multimodal on RTX 5070 Ti.
Gemma 4 E4B (Q8)
Local OKExcellentA · 48
Best for: On-device, Mobile, Chat·Pop: 82/100
Perf: est. 133 tok/s, typical error 5% · first token ~1.0s
Fits in 16 GB VRAM with room to spare. Best for on-device, mobile, chat on RTX 5070 Ti.
NVIDIA Nemotron 3.5 Bolt E4B
Local OKExcellentA · 45
Best for: Voice assistants, Tool calling, Chat·Pop: 75/100
Perf: est. 163 tok/s, typical error 5% · first token ~0.2s
Fits in 16 GB VRAM with room to spare. Best for voice assistants, tool calling, chat on RTX 5070 Ti.
Ministral 3 14B
Local OKExcellentA · 37
Best for: Chat, Reasoning, Vision·Pop: 74/100
Perf: est. 79 tok/s, typical error 15% · first token ~0.3s
Fits in 16 GB VRAM with room to spare. Best for chat, reasoning, vision on RTX 5070 Ti.
Qwen3.5 9B Instruct
Local OKExcellentA · 51
Best for: Quality, Coding, Reasoning·Pop: 86/100
Perf: est. 133 tok/s, typical error 5% · first token ~0.2s
Fits in 16 GB VRAM with room to spare. Best for quality, coding, reasoning on RTX 5070 Ti.
LFM2 24B-A2B Instruct
Local OKOKB · 3
Best for: Local AI agents, privacy-first tool calling, MCP workflows·Pop: 80/100
Perf: est. 244 tok/s, typical error 15% · first token ~0.1s
Fits in 16 GB VRAM with room to spare. Best for local ai agents, privacy-first tool calling, mcp workflows on RTX 5070 Ti.
Ornith 1.0 9B
Local OKExcellentA · 61
Best for: Agentic coding on small machines·Pop: 76/100
Perf: est. 133 tok/s, typical error 5% · first token ~0.2s
Fits in 16 GB VRAM with room to spare. Best for agentic coding on small machines on RTX 5070 Ti.
Video and image models on the RTX 5070 Ti
Runs · Stated: the linked source says a GPU with that much VRAM runs the model. Runs · Published: a dated run on the named card with that much VRAM. unknown: no source yet. Exceeds VRAM: the complete download is larger than 16 GB, so part of it has to sit in system RAM. Every VRAM tier: video · image.
Video
| Model | Fits | Quality | App |
|---|---|---|---|
| HunyuanVideo-1.5Tencent Hunyuan Community License | Runs · Stated14 GB VRAM · official runner with model offloading · HunyuanVideo-1.5 model card | #38 of 48 in text to video4,292 votessnapshot 2026-09-22 | ComfyUI |
| LTX-2.3LTX-2 Community License Agreement | Runs · Stated16 GB VRAM · distilled with FP8 cast, previews only · LTX blog: hardware for AI video models | No independent score yet | Lightricks/LTX-2 ltx-pipelines, ComfyUI |
| LTX-2.5LTX-2.x Community License Agreement | Runs · Stated12 GB VRAM · LTX Desktop on Windows, weight streaming, 40 GB RAM, 540p · LTX Desktop compatibility matrix (Lightricks) | No independent score yet | Lightricks/LTX-2 ltx-pipelines, ComfyUI |
| MiniMax H3MiniMax H3 Community License Agreement | Runs · PublishedRTX 5060 · 16 GB VRAM · ComfyUI int8 convrot with int4 text encoder, 1024x768, 10 s in 24 min · ComfyUI issue #15274 | #8 of 48 in text to video11,417 votessnapshot 2026-09-22 | stable-diffusion.cpp, ComfyUI |
| Wan2.1Apache License 2.0 | Runs · Stated13 GB VRAM · T2V 14B in Diffusers, bf16 with group offloading to system RAM · Diffusers docs: Wan pipeline | No independent score yet | ComfyUI |
| Wan2.2 A14BApache License 2.0 | Runs · PublishedRTX 5060 Ti · 16 GB VRAM · ComfyUI fp16 I2V with 4-step LoRAs, 64 GB system RAM · ComfyUI issue #10662 | #42 of 48 in text to video10,458 votessnapshot 2026-09-22 | ComfyUI |
| Wan2.2 TI2V 5BApache License 2.0 | Runs · Stated8 GB VRAM · ComfyUI native offloading · ComfyUI docs: Wan2.2Complete: Comfy-Org repackaged FP16 with FP8 scaled UMT518.1 GB download | No independent score yet | ComfyUI |
| Mochi 1 previewApache License 2.0 | unknownComplete: calcuis GGUF Q3_K_M with Q4_0 T5 and FP8 VAE7.7 GB download | #48 of 48 in text to video5,893 votessnapshot 2026-09-22 | Diffusers, ComfyUI |
Quality: blind human votes from Arena (arena.ai), CC BY 4.0, snapshot 2026-09-22. Scores are for the Arena checkpoint, not your local quantized build.
Image
| Model | Fits | Quality | App |
|---|---|---|---|
| FLUX 3 Image KleinFLUX Kommunity License v1.0 | Runs · Stated8 GB VRAM · official ComfyUI int8 files · FLUX 3 Image Klein README (Black Forest Labs) | No independent score yet | Diffusers |
| FLUX.1 [dev]FLUX.1 [dev] Non-Commercial License v1.1.1 | Runs · Stated10 GB VRAM · FP4 on RTX 50 (Blackwell) · Black Forest Labs blog: FLUX with NVIDIA | #78 of 82 in text to image49,221 votessnapshot 2026-10-07 | Diffusers, ComfyUI |
| FLUX.2 [klein]Apache 2.0 (4B), FLUX Non-Commercial License (9B) | Runs · Stated13 GB VRAM · Klein 4B · FLUX.2 [klein] 4B model cardComplete: Klein 4B BF1616.0 GB download | No independent score yet | Diffusers, stable-diffusion.cpp |
| Krea 2 TurboKrea 2 Community License Agreement v.1 | Runs · PublishedRTX 4070 · 12 GB VRAM · ComfyUI fp8, dynamic VRAM paging, 8 steps · ComfyUI issue #15365 | No independent score yet | Diffusers, stable-diffusion.cpp |
| Qwen-Image-2.1Qwen Research License Agreement | Runs · PublishedRTX 5080 · 16 GB VRAM · stable-diffusion.cpp Q8_0, 1024x1024 · stable-diffusion.cpp issue #2041 | #19 of 82 in text to image11,441 votessnapshot 2026-10-07 | Diffusers, stable-diffusion.cpp |
| SDXL Base 1.0CreativeML Open RAIL++-M License | Runs · Stated8 GB VRAM · SDXL 1.0, no app named · SDXL 1.0 announcement (Stability AI) | No independent score yet | ComfyUI, stable-diffusion.cpp |
| Stable Diffusion 3.5 LargeStability AI Community License | Runs · Stated11 GB VRAM · FP8 TensorRT build on RTX 40 and 50 · Stability AI: SD3.5 TensorRT models | #81 of 82 in text to image23,396 votessnapshot 2026-10-07 | ComfyUI |
| Z-Image-TurboApache 2.0 | Runs · Stated16 GB VRAM · stated for consumer GPUs · Z-Image-Turbo model card | #59 of 82 in text to image24,808 votessnapshot 2026-10-07 | Diffusers, stable-diffusion.cpp |
| FLUX.2 [dev]FLUX Non-Commercial License | Exceeds VRAMComplete: Q4_K_M GGUF34.8 GB download | #38 of 82 in text to image77,988 votessnapshot 2026-10-07 | Diffusers, ComfyUI, stable-diffusion.cpp |
Quality: blind human votes from Arena (arena.ai), CC BY 4.0, snapshot 2026-10-07. Scores are for the Arena checkpoint, not your local quantized build.
Can I run it on the RTX 5070 Ti?
Per-model verdict pages for this exact card: weights, KV cache by context length, and estimated tok/s, each derived from the same engine the recommendations above use.
- Qwen3 8B Q4_K_Mfits
- Qwen3 14B Q4_K_Mfits
- GPT-OSS 20B MXFP4fits
- Qwen3.8 27B Q4_K_Mtight
- Llama 3.3 70B Instruct Q4_K_Mno
- Qwen3.5 35B-A3B Instruct Q4_K_Mtight
- Gemma 4 26B-A4B Q4_K_Mtight
- Qwen3.5 4B Instruct Q4_K_Mfits
- Gemma 4 E4B Q4_K_Mfits
- Gemma 3 4B Instruct Q4_K_Mfits
- Llama 3.1 8B Instruct Q4_K_Mfits
- Qwen3.5 9B Instruct Q4_K_Mfits
- Gemma 4 12B Q4_K_Mfits
- Mistral Nemo 12B Q4_K_Mfits
- Qwen3.5 9B Instruct (Q8) Q8_0fits
- Gemma 4 12B (Q8) Q8_0fits
- LFM2 24B-A2B Instruct Q4_K_Mfits
- Qwen3.5 27B Instruct Q4_K_Mtight
- Qwen3.6 27B Q4_K_Mtight
- Gemma 4 31B Q4_K_Mtight
- Qwen3.6 35B-A3B Q4_K_Mno
- Qwen3.8 27B (Q8) Q8_0no
Similar GPUs for local AI
Speeds are estimates for Qwen3 8B Q4_K_M, typical error 5%.
Compatible model families
- QwenAlibaba Cloud: Qwen3.8 27B local flagship (Apache 2.0, vision)
- LlamaMeta: Most popular open-weight model family
- DeepSeekDeepSeek AI: V4.1 Flash: multimodal 552B MoE, API only
- MistralMistral AI: Excellent performance-per-parameter ratio
- GemmaGoogle DeepMind: Gemma 4 E4B: 4.5B effective, runs on 8GB Macs
- PhiMicrosoft: Best quality-per-gigabyte at small sizes
RTX 5070 Ti FAQ: common questions
How much VRAM does the RTX 5070 Ti have for LLMs?
The RTX 5070 Ti has 16GB GDDR7 VRAM with 896 GB/s bandwidth. About 15.5GB is usable for models. This comfortably fits all 14B models at Q4 and some 27B models at lower quantization.
What size LLM can I run on an RTX 5070 Ti?
Up to 14B parameter models at Q4 quantization fit perfectly. Popular picks: Qwen 2.5 14B, DeepSeek-R1 14B, Phi-3 14B. You can also squeeze in 27B models at Q3, though with reduced quality.
Is the RTX 5070 Ti good for local AI?
The RTX 5070 Ti is arguably the best value GPU for local AI in 2026. It matches the RTX 3090 in 8B speed (est. 125 tok/s, typical error 5% vs est. 117 tok/s), runs 14B models at est. 78 tok/s, typical error 15%, and still costs well under the RTX 5080 while delivering similar throughput.
RTX 5070 Ti vs RTX 3090 for AI: which is better?
The RTX 3090 has 24GB VRAM vs 16GB, allowing 32B models. But the 5070 Ti matches it in 8B speed (est. 125 tok/s, typical error 5% vs est. 117 tok/s) and costs less new. Choose a used 3090 for 32B models, or the 5070 Ti for 14B models with modern efficiency.
RTX 5070 Ti vs RTX 5080 for running LLMs?
Both have 16GB VRAM. The 5080 is 6% faster (est. 132 tok/s, typical error 5% vs est. 125 tok/s on 8B) but costs noticeably more. For most users, the 5070 Ti delivers 95% of the speed for less money, a clear value winner.
How fast is a 27B-class model on the RTX 5070 Ti?
The RTX 5070 Ti's 16GB of VRAM cannot fit a 32B model comfortably. The largest size class it fits is 20B MoE (3.6B active), at est. 162 tok/s, typical error 15%. The largest individual model in the catalog that fits is LFM2 24B-A2B Instruct (24B).
Related guides & benchmarks
More local AI guides
- How Much RAM (or VRAM) Do You Need for a Local LLM?The model-size-to-memory matrix: what each VRAM and RAM tier actually runs.
- Best LLM for MacBook (Apple Silicon)Unified-memory Macs run bigger models per dollar. Picks by RAM tier, M1 to M5.
- Local AI image generator on MacCompare image models by RAM and app.
- Local image generation on MacImage models and apps by Mac RAM.
Browse all NVIDIA GPUs for AI
Real-world calibration
Have this hardware? Share a measured run.
Measure a fixed Ollama reference model on your machine. This separate, opt-in contribution does not change ModelFit's third-party SWE-Bench leaderboard; results appear on the measured-speed leaderboard after the next site update.
$ npx -y @wecko-ai/modelfit bench --submit --cleanup
Want personalized recommendations?
Use our interactive wizard to compare models across Apple Silicon and NVIDIA GPUs.