FitMyLLM
INTEL· KNIGHTS

Intel Aubrey Isle

Running LLMs on the Aubrey Isle — the long read: which models fit at which quantisation, and the settings worth changing. · Or what a budget buys

VRAM
2 GB
BUDGET
BANDWIDTH
154
GB/S
MODELS Q4
47/449
10%
7B Q4 SPEED
~7
SLOW
▸ MODEL COVERAGE @ Q410% OF ALL
▸ ESTIMATED SPEED· BY MODEL SIZE @ Q4

Average speeds at Q4 quantization. Actual performance varies by model architecture and context length.

3B
1.7GB NEEDED
7B
3.9GB NEEDED
14B
7.9GB NEEDED
32B
18.0GB NEEDED
70B
39.4GB NEEDED
▸ MEASURED RIG REPORTS

We rent the machine and time every model on it: decode, VRAM peak, concurrency, watts and cost per million tokens. Yours may already be one of them — and two are free to read in full.

SEE THE REPORTS
▸ SPECIFICATIONS
VRAM
2 GB
BANDWIDTH
154 GB/s
FP16 COMPUTE
1.2 TFLOPS
TDP
300W
MEMORY
GDDR5
ARCHITECTURE
Knights
COMPUTE UNITS
32
PCIE
Gen 2 x16
47
FAST MODELS · >30 TOK/S
Real-time chat speed
47
USABLE · >10 TOK/S
Comfortable for all tasks
47
TOTAL COMPATIBLE
Fit in VRAM at Q4
▸ DON’T WANT TO BUY?

Test Aubrey Isle (or anything bigger) without committing. Pay by the second, cancel anytime.

Spin up in ~60s. Pay by the second. Cancel anytime.

Some links are affiliate links — we may earn a small commission at no extra cost to you. This helps keep FitMyLLM free and independent.

▸ COMPATIBLE MODELS· 47
S
Falcon-H1R Tiny 90M0.09B
FALCON·256K CTX· CHAT· REASONING
548
TOK/S · 27% VRAM
S
SmolLM2 135M0.135B
SMOLLM·2K CTX· CHAT
365
TOK/S · 29% VRAM
S
nomic-embed-text-v1.5 100M0.14B
EMBEDDING·8K CTX· CHAT
352
TOK/S · 29% VRAM
S
GPT-2 124M0.14B
GPT2·1K CTX· CHAT
352
TOK/S · 29% VRAM
S
SmolVLM 256M0.256B
SMOLLM·8K CTX· CHAT· VISION
192
TOK/S · 32% VRAM
S
Gemma 3 270M0.27B
GEMMA·32K CTX· CHAT
183
TOK/S · 33% VRAM
S
Snowflake Arctic Embed M v2.00.305B
EMBEDDING·8K CTX· EMBEDDING· MULTILINGUAL
162
TOK/S · 34% VRAM
S
bge-large-en-v1.5 335M0.335B
EMBEDDING·1K CTX· CHAT
147
TOK/S · 35% VRAM
S
mxbai-embed-large-v10.335B
EMBEDDING·1K CTX· EMBEDDING
147
TOK/S · 35% VRAM
S
Snowflake Arctic Embed L0.335B
EMBEDDING·1K CTX· EMBEDDING
147
TOK/S · 35% VRAM
S
LFM2 350M0.35B
LFM·125K CTX· CHAT· TOOL_USE
141
TOK/S · 35% VRAM
S
SmolLM2 360M0.36B
SMOLLM·8K CTX· CHAT
137
TOK/S · 35% VRAM
S
GPT-2 Medium 345M0.38B
GPT2·1K CTX· CHAT
130
TOK/S · 36% VRAM
S
Qwen 2.5 0.5B0.5B
QWEN·32K CTX· CHAT
99
TOK/S · 40% VRAM
S
SmolVLM 500M0.5B
SMOLLM·8K CTX· CHAT· VISION
99
TOK/S · 40% VRAM
S
Falcon-H1 0.5B0.52B
FALCON·128K CTX· CHAT
95
TOK/S · 40% VRAM
S
BGE-M30.568B
EMBEDDING·8K CTX· EMBEDDING
87
TOK/S · 42% VRAM
S
Snowflake Arctic Embed L v2.00.568B
EMBEDDING·8K CTX· EMBEDDING· MULTILINGUAL
87
TOK/S · 42% VRAM
S
Qwen3-Embedding 0.6B0.6B
EMBEDDING·32K CTX· EMBEDDING· MULTILINGUAL
82
TOK/S · 43% VRAM
S
Falcon-H1R Tiny 0.6B0.6B
FALCON·32K CTX· CHAT· REASONING
82
TOK/S · 43% VRAM
S
Falcon Perception 0.6B0.6B
FALCON·4K CTX· VISION
82
TOK/S · 43% VRAM
S
Qwen 1.5 0.5B0.62B
QWEN·32K CTX· CHAT
79
TOK/S · 43% VRAM
S
LFM2 700M0.74B
LFM·125K CTX· CHAT· TOOL_USE
67
TOK/S · 47% VRAM
S
Qwen3 0.6B0.75B
QWEN·32K CTX· CHAT· REASONING
66
TOK/S · 47% VRAM
S
GPT-2 Large 774M0.81B
GPT2·1K CTX· CHAT
61
TOK/S · 49% VRAM
A
Qwen 3.5 0.8B0.87B
QWEN·256K CTX· CHAT· CODING· MULTILINGUAL
57
TOK/S · 51% VRAM
A
InternVL3 1B0.94B
OTHER·32K CTX· CHAT· VISION
52
TOK/S · 53% VRAM
A
MiniCPM5 1B1.08B
MINICPM·128K CTX· CHAT· REASONING· MULTILINGUAL
46
TOK/S · 57% VRAM
A
TinyLlama 1.1B1.1B
LLAMA·2K CTX· CHAT
45
TOK/S · 58% VRAM
A
LFM2.5-1.2B-Thinking1.2B
LFM·122K CTX· CHAT· REASONING· TOOL_USE
41
TOK/S · 61% VRAM
A
Llama-3.2-1B1.2B
LLAMA·4K CTX· CHAT
41
TOK/S · 61% VRAM
A
LFM2 1.2B1.2B
LFM·125K CTX· CHAT· TOOL_USE· MULTILINGUAL
41
TOK/S · 61% VRAM
A
Zamba2 1.2B1.2B
OTHER·4K CTX· CHAT
41
TOK/S · 61% VRAM
B
EXAONE-4.0-1.2B1.3B
EXAONE·64K CTX· CHAT
38
TOK/S · 64% VRAM
B
OPT 1.3B1.3B
OPT·2K CTX· CHAT
38
TOK/S · 64% VRAM
B
MiniCPM-V 4.61.3B
OTHER·256K CTX· CHAT· VISION
38
TOK/S · 64% VRAM
B
DeepSeek Coder 1.3B1.35B
DEEPSEEK·16K CTX· CODING
37
TOK/S · 66% VRAM
B
Phi-1 1.3B1.42B
PHI·2K CTX· CODING
35
TOK/S · 68% VRAM
B
Phi-1.5 1.3B1.42B
PHI·2K CTX· CHAT· CODING
35
TOK/S · 68% VRAM
B
Qwen2.5-Coder-1.5B1.5B
QWEN·32K CTX· CHAT· TOOL_USE· CODING
33
TOK/S · 70% VRAM
▸ NEXT STEP

Get personalized recommendations.

See ranked models with benchmark scores, run commands, and precise speed estimates for your Aubrey Isle.

WHAT THIS CARD IS WORTH

Aubrey Isle holds 47 of the models in our catalogue and is, in practice, a Q4_K_M card — the largest it takes is Moondream2 1.9B at Q4_K_M.

TOKENS/SEC PER $100
8B at Q4_K_M, so cards compare like for like
VRAM PER $100
what memory costs on this card
THE BIGGEST IT TAKES
Moondream2 1.9B1.9B · Q4_K_M1.7 GB24 tok/sEST
Falcon3-1B1.67B · Q5_K_M1.8 GB23 tok/sEST
LFM2-VL 1.6B1.6B · Q5_K_M1.7 GB24 tok/sEST
Falcon-H1 1.5B1.55B · Q6_K1.7 GB22 tok/sEST
Qwen2.5-Coder-1.5B1.5B · Q6_K1.6 GB23 tok/sEST
Qwen2 Math 1.5B1.5B · Q6_K1.6 GB23 tok/sEST
Qwen 2.5 1.5B1.5B · Q6_K1.6 GB23 tok/sEST
Stella en 1.5B v51.5B · Q6_K1.6 GB23 tok/sEST
AGAINST CARDS OF SIMILAR MEMORY
M2 (8GB)5 GB$5992.8 tok/s per $100
M3 (8GB)5 GB$5992.8 tok/s per $100
M1 (8GB)5 GB$4992.2 tok/s per $100
RTX 2060 6GB6 GB$15026.7 tok/s per $100
Aubrey Isle2 GB

Neighbours in memory rather than in price: memory decides whether a card can do the job at all, so two cards of the same size at different prices is the comparison you are making. Ordered by memory, then bandwidth — not by the value column, which is worked out from bandwidth and price alone and therefore rewards a cheap card whatever its software stack does to that bandwidth in practice. Read it as one input, not as a ranking.

▸ DEVICE UNDER TEST

Intel Aubrey Isle — 2 GB VRAM.

AUBREY ISLE SPEC
BRAND
Intel
VRAM
2 GB GDDR5
BANDWIDTH
154 GB/s
FP16 COMPUTE
1.2 TFLOPS
FP32 COMPUTE
1.2 TFLOPS
TDP
300 W
ARCHITECTURE
Knights
▸ AI CAPABILITY
47/ 449 models @ Q4

With 2 GB VRAM and 154 GB/s bandwidth, this GPU handles models up to 1.61B parameters.

Speed ≈ bandwidth / model_size × efficiency. A 7B model at Q4 runs at ~18 tok/s.

§ 01TOP MODELS FOR AUBREY ISLE
47 FIT · SHOWING 20
MODELSIZEVRAM Q4TOK/SAVG
GPT-2 XL 1.5B1.61B1.5 GB315.1
stablelm-2-1_6b1.6B1.5 GB319.5
Falcon-H1 1.5B1.55B1.4 GB3243.8
Qwen2.5-Coder-1.5B1.5B1.4 GB3319.6
Qwen2 Math 1.5B1.5B1.4 GB3319.6
Qwen 2.5 1.5B1.5B1.4 GB3330.2
Yi Coder 1.5B1.5B1.4 GB3314.6
Stella en 1.5B v51.5B1.4 GB33
Phi-1 1.3B1.42B1.4 GB357.2
Phi-1.5 1.3B1.42B1.4 GB357.2
DeepSeek Coder 1.3B1.35B1.3 GB3716.8
EXAONE-4.0-1.2B1.3B1.3 GB3818.9
OPT 1.3B1.3B1.3 GB385.3
MiniCPM-V 4.61.3B1.3 GB3821.5
LFM2.5-1.2B-Thinking1.2B1.2 GB4119.6
Llama-3.2-1B1.2B1.2 GB4110.1
LFM2 1.2B1.2B1.2 GB4115.6
Zamba2 1.2B1.2B1.2 GB4141.5
TinyLlama 1.1B1.1B1.2 GB4513.6
MiniCPM5 1B1.08B1.1 GB4626.5