FitMyLLM
▸ ARM· VALHALL

ARM Mali-G78 MP14

Running LLMs on the ARM Mali-G78 MP14 — the long read: which models fit at which quantisation, and the settings worth changing. · Or what a budget buys

VRAM
8 GB
ENTRY-LEVEL
BANDWIDTH
44
GB/S
MODELS Q4
201/449
45%
7B Q4 SPEED
~6
SLOW
▸ MODEL COVERAGE @ Q445% OF ALL
▸ ESTIMATED SPEED· BY MODEL SIZE @ Q4

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

3B
~13
TOK/S
7B
~6
TOK/S
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
8 GB
BANDWIDTH
44 GB/s
FP16 COMPUTE
0.8 TFLOPS
TDP
6W
MEMORY
—
ARCHITECTURE
Valhall
36
FAST MODELS · >30 TOK/S
Real-time chat speed
101
USABLE · >10 TOK/S
Comfortable for all tasks
201
TOTAL COMPATIBLE
Fit in VRAM at Q4
▸ DON’T WANT TO BUY?

Test ARM Mali-G78 MP14 (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· 201
S
Falcon-H1R Tiny 90M0.09B
FALCON·256K CTX· CHAT· REASONING
435
TOK/S · 7% VRAM
›
S
SmolLM2 135M0.135B
SMOLLM·2K CTX· CHAT
290
TOK/S · 7% VRAM
›
S
nomic-embed-text-v1.5 100M0.14B
EMBEDDING·8K CTX· CHAT
279
TOK/S · 7% VRAM
›
S
GPT-2 124M0.14B
GPT2·1K CTX· CHAT
279
TOK/S · 7% VRAM
›
S
SmolVLM 256M0.256B
SMOLLM·8K CTX· CHAT· VISION
153
TOK/S · 8% VRAM
›
S
Gemma 3 270M0.27B
GEMMA·32K CTX· CHAT
145
TOK/S · 8% VRAM
›
S
Snowflake Arctic Embed M v2.00.305B
EMBEDDING·8K CTX· EMBEDDING· MULTILINGUAL
128
TOK/S · 8% VRAM
›
S
bge-large-en-v1.5 335M0.335B
EMBEDDING·1K CTX· CHAT
117
TOK/S · 9% VRAM
›
S
mxbai-embed-large-v10.335B
EMBEDDING·1K CTX· EMBEDDING
117
TOK/S · 9% VRAM
›
S
Snowflake Arctic Embed L0.335B
EMBEDDING·1K CTX· EMBEDDING
117
TOK/S · 9% VRAM
›
S
LFM2 350M0.35B
LFM·125K CTX· CHAT· TOOL_USE
112
TOK/S · 9% VRAM
›
S
SmolLM2 360M0.36B
SMOLLM·8K CTX· CHAT
109
TOK/S · 9% VRAM
›
S
GPT-2 Medium 345M0.38B
GPT2·1K CTX· CHAT
103
TOK/S · 9% VRAM
›
S
Qwen 2.5 0.5B0.5B
QWEN·32K CTX· CHAT
78
TOK/S · 10% VRAM
›
S
SmolVLM 500M0.5B
SMOLLM·8K CTX· CHAT· VISION
78
TOK/S · 10% VRAM
›
S
Falcon-H1 0.5B0.52B
FALCON·128K CTX· CHAT
75
TOK/S · 10% VRAM
›
S
BGE-M30.568B
EMBEDDING·8K CTX· EMBEDDING
69
TOK/S · 10% VRAM
›
S
Snowflake Arctic Embed L v2.00.568B
EMBEDDING·8K CTX· EMBEDDING· MULTILINGUAL
69
TOK/S · 10% VRAM
›
S
Qwen3-Embedding 0.6B0.6B
EMBEDDING·32K CTX· EMBEDDING· MULTILINGUAL
65
TOK/S · 11% VRAM
›
S
Falcon-H1R Tiny 0.6B0.6B
FALCON·32K CTX· CHAT· REASONING
65
TOK/S · 11% VRAM
›
S
Falcon Perception 0.6B0.6B
FALCON·4K CTX· VISION
65
TOK/S · 11% VRAM
›
S
Qwen 1.5 0.5B0.62B
QWEN·32K CTX· CHAT
63
TOK/S · 11% VRAM
›
A
LFM2 700M0.74B
LFM·125K CTX· CHAT· TOOL_USE
53
TOK/S · 12% VRAM
›
A
Qwen3 0.6B0.75B
QWEN·32K CTX· CHAT· REASONING
52
TOK/S · 12% VRAM
›
A
GPT-2 Large 774M0.81B
GPT2·1K CTX· CHAT
48
TOK/S · 12% VRAM
›
A
Qwen 3.5 0.8B0.87B
QWEN·256K CTX· CHAT· CODING· MULTILINGUAL
45
TOK/S · 13% VRAM
›
A
InternVL3 1B0.94B
OTHER·32K CTX· CHAT· VISION
42
TOK/S · 13% VRAM
›
B
TinyLlama 1.1B1.1B
LLAMA·2K CTX· CHAT
36
TOK/S · 15% VRAM
›
B
MiniCPM5 1B1.08B
MINICPM·128K CTX· CHAT· REASONING· MULTILINGUAL
36
TOK/S · 14% VRAM
›
B
LFM2.5-1.2B-Thinking1.2B
LFM·122K CTX· CHAT· REASONING· TOOL_USE
33
TOK/S · 15% VRAM
›
B
Llama-3.2-1B1.2B
LLAMA·4K CTX· CHAT
33
TOK/S · 15% VRAM
›
B
LFM2 1.2B1.2B
LFM·125K CTX· CHAT· TOOL_USE· MULTILINGUAL
33
TOK/S · 15% VRAM
›
B
Zamba2 1.2B1.2B
OTHER·4K CTX· CHAT
33
TOK/S · 15% VRAM
›
B
EXAONE-4.0-1.2B1.3B
EXAONE·64K CTX· CHAT
30
TOK/S · 16% VRAM
›
B
OPT 1.3B1.3B
OPT·2K CTX· CHAT
30
TOK/S · 16% VRAM
›
B
MiniCPM-V 4.61.3B
OTHER·256K CTX· CHAT· VISION
30
TOK/S · 16% VRAM
›
B
DeepSeek Coder 1.3B1.35B
DEEPSEEK·16K CTX· CODING
29
TOK/S · 16% VRAM
›
B
Phi-1 1.3B1.42B
PHI·2K CTX· CODING
28
TOK/S · 17% VRAM
›
B
Phi-1.5 1.3B1.42B
PHI·2K CTX· CHAT· CODING
28
TOK/S · 17% VRAM
›
B
LFM2 8B A1B8.3BMoE
LFM·125K CTX· CHAT· TOOL_USE· MULTILINGUAL
26
TOK/S · 70% VRAM
›
▸ NEXT STEP

Get personalized recommendations.

See ranked models with benchmark scores, run commands, and precise speed estimates for your ARM Mali-G78 MP14.

WHAT THIS CARD IS WORTH

ARM Mali-G78 MP14 holds 201 of the models in our catalogue and is, in practice, a IQ4_XS card — the largest it takes is Falcon3-10B at IQ4_XS.

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
Falcon3-10B10.3B · IQ4_XS7.1 GB6 tok/sEST
Bamba 9B v29.78B · Q4_K_M7.2 GB6 tok/sEST
RecurrentGemma 9B9.63B · Q4_K_M6.8 GB6 tok/sEST
glm-4-9b9.4B · Q4_K_M6.6 GB6 tok/sEST
Yi 1.5 9B9B · Q4_K_M6.5 GB6 tok/sEST
Yi Coder 9B9B · Q4_K_M6.5 GB6 tok/sEST
Ministral 3 8B8.92B · Q4_K_M6.6 GB6 tok/sEST
Ministral 3 8B Reasoning8.92B · Q4_K_M6.6 GB6 tok/sEST
AGAINST CARDS OF SIMILAR MEMORY
M3 Pro (18GB)12 GB$1,5991.5 tok/s per $100
M1 Pro (16GB)11 GB$9992.9 tok/s per $100
M2 Pro (16GB)11 GB$1,2992.2 tok/s per $100
M4 (16GB)11 GB$4994.0 tok/s per $100
ARM Mali-G78 MP148 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

ARM Mali-G78 MP14 — 8 GB VRAM.

▸ ARM MALI-G78 MP14 SPEC
BRAND
Apple
VRAM
8 GB
BANDWIDTH
44 GB/s
FP16 COMPUTE
0.8 TFLOPS
FP32 COMPUTE
0.4 TFLOPS
TDP
6 W
ARCHITECTURE
Valhall
▸ AI CAPABILITY
201/ 449 models @ Q4

With 8 GB VRAM and 44 GB/s bandwidth, this GPU handles models up to 9.63B parameters.

Speed ≈ bandwidth / model_size × efficiency. A 7B model at Q4 runs at ~5 tok/s.

RENT IT — LIVE PRICES
Checking what the clouds are charging…
§ 01TOP MODELS FOR ARM MALI-G78 MP14
201 FIT · SHOWING 20
MODELSIZEVRAM Q4TOK/SAVG
RecurrentGemma 9B9.63B6.4 GB435.0
glm-4-9b9.4B6.2 GB420.5
gemma-2-9b9.2B6.1 GB430.2
Yi 1.5 9B9B6.0 GB430.3
Yi Coder 9B9B6.0 GB435.8
Ministral 3 8B8.92B5.9 GB425.7
Ministral 3 8B Reasoning8.92B5.9 GB4—
NVIDIA-Nemotron-Nano-9B-v28.9B5.9 GB444.2
InternLM3 8B Instruct8.8B5.9 GB438.7
Gemma 1 7B8.54B5.7 GB524.7
CodeGemma 7B8.54B5.7 GB540.2
LFM2 8B A1B8.3B5.6 GB2624.3
Seed-Coder 8B Instruct8.25B5.5 GB534.1
Seed-Coder 8B Reasoning8.25B5.5 GB532.9
DeepSeek R1-0528 Qwen3 8B8.2B5.5 GB536.3
Qwen3-8B8.2B5.5 GB543.3
Granite 3.0 8B8.17B5.5 GB536.4
Granite 3.1 8B8.17B5.5 GB538.6
Command-R7B8.03B5.4 GB535.3
Aya Expanse 8B8B5.4 GB527.8