dexiio

Best local LLMs for the CPU only / integrated graphics (2026)

The CPU only / integrated graphics has 0 GB of VRAM and 0 GB/s of memory bandwidth. That fits 0 of our 30 tracked models entirely on the GPU at Q4_K_M and 32K context. Every figure below is computed from weights + KV cache + overhead, not guessed. Open this GPU in the calculator →

All figures assume an f16 KV cache, a 0.6 GB display reserve on the GPU, and 64 GB of DDR5 system RAM for the offload tiers. Tune these in the calculator.

Fit grid by context length

Model8K16K32K64K128K
Llama 3.1 8BQ8_0Q8_0Q8_0Q8_0Q8_0
Qwen3.5 9BQ8_0Q8_0Q8_0Q8_0Q8_0
Qwen3.5 4BQ8_0Q8_0Q8_0Q8_0Q8_0
Gemma 3 4BQ8_0Q8_0Q8_0Q8_0Q8_0
Mistral Nemo 12BQ8_0Q8_0Q8_0Q8_0Q8_0
Phi 4 14BQ8_0Q8_0———
Gemma 4 12BQ8_0Q8_0Q8_0Q8_0Q8_0
Qwen3 14BQ8_0Q8_0Q8_0——
Gemma 3 12BQ8_0Q8_0Q8_0Q8_0Q8_0
DeepSeek R1 Distill Qwen 14BQ8_0Q8_0Q8_0Q8_0Q8_0
GPT OSS 20BQ8_0Q8_0Q8_0Q8_0Q8_0
Rocinante 12BQ8_0Q8_0Q8_0Q8_0Q8_0
Qwen3.5 27BQ8_0Q8_0Q8_0Q8_0Q5_K_M
Qwen3.5 35B A3BQ8_0Q8_0Q8_0Q8_0Q8_0
Qwen3 Coder 30B A3BQ8_0Q8_0Q8_0Q8_0Q8_0
Gemma 3 27BQ8_0Q8_0Q8_0Q8_0Q8_0
Gemma 4 26B A4BQ8_0Q8_0Q8_0Q8_0Q5_K_M
Mistral Small 3.2 24BQ8_0Q8_0Q8_0Q8_0Q8_0
DeepSeek R1 Distill Qwen 32BQ8_0Q8_0Q8_0Q8_0Q4_K_M
Cydonia 24BQ8_0Q8_0Q8_0Q8_0Q8_0
Llama 3.3 70BQ5_K_MQ4_K_MQ4_K_M——
Llama 4 ScoutQ3_K_M————
GPT OSS 120B—————
GLM 4.5 Air—————
Qwen3.5 122B A10B—————
Qwen3.5 397B A17B—————
DeepSeek R1—————
Mistral Large 3—————
Kimi K2.5—————
Anubis 70BQ5_K_MQ4_K_MQ4_K_M——

Fits on GPUExpert offloadPartial offloadCPU only

Top pick per use case

Coding · 32K

Qwen3 Coder 30B A3B Q8_0

CPU only · ≈ 14 tok/s

Purpose-built agentic coder. Best local fill-in-the-middle and tool-calling under 70B; useless at small talk.

Roleplay & writing · 16K

Rocinante 12B Q8_0

CPU only · ≈ 4 tok/s

The budget roleplay king. Lowest slop-per-token of anything under 24 GB; the community keeps it alive for a reason.

Summarization · 32K

Qwen3.5 27B Q8_0

CPU only · ≈ 2 tok/s

The dense 24GB workhorse. If you want one model on a 3090 and no surprises, it's this.

RAG & documents · 16K

Qwen3.5 35B A3B Q8_0

CPU only · ≈ 13 tok/s

The meta pick, full stop. Near-dense-30B quality at 3B-active speed, and expert offload puts it on 8 GB cards.

Vision / image input · 16K

Gemma 3 27B Q8_0

CPU only · ≈ 2 tok/s

Sliding-window attention keeps long context cheap, and the vision stack still beats most of 2026's newcomers.

Almost fits

These models can't run well on 0 GB at 32K: Llama 3.1 8B, Qwen3.5 9B, Qwen3.5 4B, Gemma 3 4B, Mistral Nemo 12B, Phi 4 14B and 24 more.

What an upgrade unlocks

Stepping up to a Apple M2 Ultra (128GB) (96 GB) unlocks 25 more models on GPU or expert offload at 32K, including Llama 3.1 8B, Qwen3.5 9B, Qwen3.5 4B.

Frequently asked questions

What is the best local LLM for a CPU only / integrated graphics in 2026?

Qwen3 Coder 30B A3B is our top overall pick on the CPU only / integrated graphics: Purpose-built agentic coder. Best local fill-in-the-middle and tool-calling under 70B; useless at small talk.

How many local LLMs fit in 0 GB of VRAM?

At Q4_K_M quantization and 32K context, 0 of our 30 tracked models fit entirely in the CPU only / integrated graphics's 0 GB of VRAM, and 0 more MoE models run via expert offload with enough system RAM.

Can a CPU only / integrated graphics run a 70B model like Llama 3.3?

Yes — Llama 3.3 70B runs on the CPU only / integrated graphics as "CPU only" at Q4_K_M, around 1 tokens/sec.

Can a CPU only / integrated graphics run DeepSeek R1?

Not the full 671B model — its Q2_K weights alone exceed 200 GB. The R1-Distill-Qwen 14B/32B models are the practical local alternative on this card.

How much VRAM do I need for 32K context?

The KV cache is separate from the weights and grows linearly with context. For a typical 8-14B dense model at 32K and f16 KV, budget 2-4 GB extra on top of the weights; MLA models like DeepSeek R1 need far less, and quantized KV (q8_0) halves it.

Related guides