Phi 3 Small 8k Instruct — Hardware Requirements & GPU Compatibility
ChatCode
Phi-3-Small-8K-Instruct is Microsoft's 7-billion-parameter instruction-tuned chat model from the Phi-3 family, trained on a mix of synthetic data and heavily filtered public web text chosen for high reasoning density, then post-trained with supervised fine-tuning and direct preference optimization. It targets memory- and compute-constrained, latency-sensitive deployments while still emphasizing strong code, math, and logical reasoning, and Microsoft reports it holds up well against same-size and next-size-up models on common-sense, language, math, code, and long-context benchmarks. It sits alongside Mini and Medium-sized Phi-3 variants, plus a 128K-context Small sibling. At around 7 billion parameters it runs comfortably on a single consumer GPU. Context length is 8,192 tokens, as the name indicates. It is released under the MIT license, permitting unrestricted commercial and research use, and was published in May 2024.
Runs well on a 24 GB GPU
Runs from 16 GB (BF16 · 4K)Best on 24 GB (BF16 · ~43 tok/s · 8K)Macs from 24 GB (BF16 · ~14 tok/s)
MIT
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Phi 3 Small 8k Instruct
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
vllm serve microsoft/Phi-3-small-8k-instruct --max-model-len 8192vLLM repo with its own quantisation
Download: Weights on Hugging Face
Ways to runOther ways
each opens step-by-stepRun in cloud
Runs on RTX 3090 (24 GB) (7 GB headroom) · BF16
- Generation speed
- ~40 tok/s
- generation speed
- Cost per 1M output tokens
- $1.19
- per 1M output tokens
How Much VRAM Does Phi 3 Small 8k Instruct Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 15.3 GB | — | 16-bit · bfloat — Brain floating point 16 — preferred for training | Fits: 8K context · ~43 tok/s |
Selected BF16
Brain floating point 1616.2GBat 8K context
- Weights
- 14.8 GB
- KV cache
- 1.1 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 7.8 GB free · ~43 tok/s
≥50 tok/s needs ≥1.18 TB/s · RTX 5090 ~76
Which Hardware Can Run Phi 3 Small 8k Instruct?
BF16 · 15.3 GBPhi 3 Small 8k Instruct (BF16) requires 15.3 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 20+ GB is recommended. Using the full 8K context window can add up to 0.8 GB, bringing total usage to 16.2 GB. 26 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 5080. 122 devices with unified memory can run Phi 3 Small 8k Instruct, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Mini M5 Pro (24 GB).
20 GB
fits · 3.8 GB free · up to 8K ctx
1 GPU
RX 7900 XT~31Simulate output on RX 7900 XT32–40 GB
fits · 15.8 GB free · up to 8K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~76Simulate output on RTX 5090M5 Max36 GB~21Simulate output on Mac Studio M5 Max (36 GB)M532 GB~7Simulate output on MacBook Air 13" M5 (32 GB)Orin~9Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~6Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at BF16
How good is Phi 3 Small 8k Instruct?
llmrun Score blends public benchmark boards into one number; 100 = the average open model.
Not enough data yet for an llmrun Score.
A Score needs results on enough independent boards. How the Composite Score is built
Benchmarks
The 6 boards that weigh most in the llmrun Score, of 8 with a result for this model; badges show who ran the test.
- LMArena Text
LMArena · Oct 2026Arena
Arena1171.1
#160/188 open
- MMLU
Epoch AI · Apr 2024Lab-reported
Lab-reported75.7%
#20/85 open
- BIG-Bench Hard
Epoch AI · Apr 2024Lab-reported
Lab-reported79.1%
#5/38 open
- HellaSwag
Epoch AI · Apr 2024Lab-reported
Lab-reported77.0%
#20/47 open
- ARC Challenge
Epoch AI · Apr 2024Lab-reported
Lab-reported90.7%
#6/56 open
- WinoGrande
Epoch AI · Apr 2024Lab-reported
Lab-reported81.5%
#9/51 open
Alternatives to Phi 3 Small 8k Instruct
Phi 3 family · 15 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Ling 3.0 TinyInclusion AI | — | 5.5 GB | ~364 | same size |
| Olmo 3 7B InstructAllen AI | — | 8.4 GB | ~114 | same size |
| Gemma 3n E4B ITGoogle | — | 7.7 GB | ~127 | same size |
| Falcon H1 7B InstructTII UAE | — | 6.3 GB | ~130 | same size |
| InternVL3 8BOpenGVLab | — | 7.8 GB | ~125 | same size |
| DeepSeek R1 Distill Qwen 7BDeepSeek | — | 5.8 GB | ~131 | same size |
Specifications
- Publisher
- Microsoft
- Family
- Phi 3
- Parameters
- 7.4B
- Architecture
- Phi3SmallForCausalLM
- Context Length
- 8,192 tokens
- Vocabulary Size
- 100,352
- Release Date
- 2024-05-07
- License
- MIT
Frequently Asked Questions
How much VRAM does Phi 3 Small 8k Instruct need?
Phi 3 Small 8k Instruct requires 15.3 GB of VRAM at BF16. Full 8K context adds up to 0.8 GB (16.2 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 7.4B × 16 bits ÷ 8 = 14.8 GB
KV Cache + Overhead ≈ 0.5 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 1.4 GB (at full 8K context)
VRAM usage by quantization
Can I run Phi 3 Small 8k Instruct on a Mac?
Yes — Mac Mini M5 Pro (24 GB) and 106 other Macs can run Phi 3 Small 8k Instruct. Apple Silicon uses unified memory, so the model shares RAM with the system. At BF16 you need at least 15.3 GB of usable unified memory (RAM minus macOS overhead).
Can I run Phi 3 Small 8k Instruct locally?
Yes — Phi 3 Small 8k Instruct can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Phi 3 Small 8k Instruct?
At BF16, Phi 3 Small 8k Instruct can reach ~313 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~43 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 15.35 × 0.65 = ~339 tok/s
Estimated speed at BF16 (15.3 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Phi 3 Small 8k Instruct?
At BF16, the download is about 14.78 GB.
Which GPUs can run Phi 3 Small 8k Instruct?
26 consumer GPUs can run Phi 3 Small 8k Instruct at BF16. Top options include AMD Radeon RX 7900 XTX, NVIDIA GeForce RTX 3090, NVIDIA GeForce RTX 3090 Ti, AMD Radeon RX 6800. 7 GPUs have plenty of headroom for comfortable inference.
Which devices can run Phi 3 Small 8k Instruct?
122 devices with unified memory can run Phi 3 Small 8k Instruct at BF16 (15.3 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.