MathGenie · text

MathCoder2-CodeLlama-7B

MathGenie/MathCoder2-CodeLlama-7B

MathCoder2-CodeLlama-7B at Q4_K_M is exactly 4,081,096,640 bytes (3.80 GiB / 4.08 GB) — an effective 4.845 bits per weight, not the nominal 4. Its KV cache at 32K is 16.00 GiB.

From the file· summed from 1 file(s)From the file· KV per layer
Parameters
6.7B
Architecture
llama
32 layers
Context
16,384
native (config.json)
License
apache-2.0

Shipped quantizations

● exact bytes, summed from published files
QuantSize●Exact bytes●Effective bpw●Tensors●Publisher
IQ2_M2.20 GiB2,359,825,3442.802—bartowski
Q2_K2.36 GiB2,532,940,7363.007—bartowski
Q2_K_L2.48 GiB2,661,004,7363.159—bartowski
IQ3_XS2.60 GiB2,796,606,9123.320—bartowski
Q3_K_S2.75 GiB2,948,388,2883.500—bartowski
IQ3_M2.90 GiB3,114,948,0323.698—bartowski
Q3_K_M3.07 GiB3,298,088,3843.916—bartowski
Q3_K_L3.35 GiB3,597,194,6884.271—bartowski
IQ4_XS3.37 GiB3,619,426,2404.297—bartowski
Q4_03.57 GiB3,837,171,6484.556—bartowski
Q4_K_S3.59 GiB3,856,832,4484.579—bartowski
Q4_K_M3.80 GiB4,081,096,6404.845—bartowski
Q4_K_L3.89 GiB4,178,425,2804.961—bartowski
Q5_K_S4.33 GiB4,651,792,3205.523—bartowski
Q5_K_M4.45 GiB4,783,257,5365.679—bartowski
Q5_K_L4.53 GiB4,864,193,9845.775—bartowski
Q6_K5.15 GiB5,529,303,4886.564—bartowski
Q6_K_L5.21 GiB5,592,823,2326.640—bartowski
Q8_06.67 GiB7,161,230,7848.502—bartowski
F1612.55 GiB13,478,368,41616.002—bartowski

KV cache by context

computed per layer
ContextKV cache (f16)●Flat formulaOverstated byFull / windowed / recurrent
4,0962.00 GiB2.00 GiB—32 / 0 / 0
8,1924.00 GiB4.00 GiB—32 / 0 / 0
16,3848.00 GiB8.00 GiB—32 / 0 / 0
32,76816.00 GiB16.00 GiB—32 / 0 / 0
65,53632.00 GiB32.00 GiB—32 / 0 / 0
131,07264.00 GiB64.00 GiB—32 / 0 / 0

Compare with

same modality, comparable size
vs gemma-4-E4B-itvs Qwen3-4Bvs Qwen3-8Bvs HyperCLOVAX-SEED-Text-Instruct-1.5Bvs Llama-3.2-1B-Instructvs llama-3-youko-8bAll alternatives →Will it fit on my card? →

Will it run on your card?

full quant x context sweep
By VRAM:8GB12GB16GB24GB32GB48GB
Radeon RX 6500 XT 4GBGeForce RTX 3050 6GBGeForce RTX 5050 8GBGeForce RTX 3080 10GBGeForce RTX 2080 Ti 11GBGeForce RTX 5070 12GBGeForce RTX 5060 Ti 16GBApple M3 Pro 18GBGeForce RTX 3080 Ti 20GBGeForce RTX 5090 D V2 24GBGeForce RTX 5090 D 32GBApple M5 Max 36GBApple M5 Pro 48GBApple M5 Max 64GBApple M3 Ultra 96GBApple M5 Max 128GBApple M2 Ultra 192GBApple M3 Ultra 256GBApple M3 Ultra 512GB

Why other calculators give a different number

A parameters × bits ÷ 8 estimate puts Q4_K_M at roughly 3.53 GiB. The real file is 3.80 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

from config.json
Layers
32
Attention heads
32
KV heads
32
Head dim
128
Hidden size
4096
Vocab
32,016
Sliding window
none
SWA period
—
MLA
no
Experts
—
Experts per token
—
use_sliding_window
—

Questions people ask

How much VRAM does MathCoder2-CodeLlama-7B need?
Q4_K_M is exactly 4,081,096,640 bytes (3.80 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is MathCoder2-CodeLlama-7B's KV cache?
16.00 GiB at 32K context with an f16 cache, computed per layer. Quantizing the cache to q8_0 roughly halves it, which is often the difference between a context length fitting and not.
Which quantization of MathCoder2-CodeLlama-7B should I use?
Q4_K_M is the usual default. Pick the largest quantization that fits your card at the context you actually need — the table above gives exact sizes for every one published.