FitMyLLM
▸ NVIDIA· ADA LOVELACE

NVIDIA GeForce RTX 4070 Ti

Running LLMs on the GeForce RTX 4070 Ti — the long read: which models fit at which quantisation, and the settings worth changing. · Or what else $799 buys

VRAM
12 GB
ENTRY-LEVEL
BANDWIDTH
504
GB/S
MODELS Q4
248/449
55%
7B Q4 SPEED
~58
BLAZING
▸ MODEL COVERAGE @ Q455% OF ALL
▸ ESTIMATED SPEED· BY MODEL SIZE @ Q4

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

3B
~134
TOK/S
7B
~58
TOK/S
14B
~29
TOK/S
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
12 GB
BANDWIDTH
504 GB/s
FP16 COMPUTE
40.1 TFLOPS
TDP
285W
MEMORY
GDDR6X
ARCHITECTURE
Ada Lovelace
CUDA CORES
7,680
TENSOR CORES
240
PCIE
Gen 4 x16
MSRP
$799
237
FAST MODELS · >30 TOK/S
Real-time chat speed
248
USABLE · >10 TOK/S
Comfortable for all tasks
248
TOTAL COMPATIBLE
Fit in VRAM at Q4
▸ WHERE TO BUY
▸Amazon ↗▸Newegg ↗

MSRP $799

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

▸ DON’T WANT TO BUY?

Test GeForce RTX 4070 Ti (or anything bigger) without committing. Pay by the second, cancel anytime.

Rent on Vast.ai →Or RunPod →

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· 248
S
Falcon-H1R Tiny 90M0.09B
FALCON·256K CTX· CHAT· REASONING
4480
TOK/S · 5% VRAM
›
S
SmolLM2 135M0.135B
SMOLLM·2K CTX· CHAT
2987
TOK/S · 5% VRAM
›
S
nomic-embed-text-v1.5 100M0.14B
EMBEDDING·8K CTX· CHAT
2880
TOK/S · 5% VRAM
›
S
GPT-2 124M0.14B
GPT2·1K CTX· CHAT
2880
TOK/S · 5% VRAM
›
S
SmolVLM 256M0.256B
SMOLLM·8K CTX· CHAT· VISION
1575
TOK/S · 5% VRAM
›
S
Gemma 3 270M0.27B
GEMMA·32K CTX· CHAT
1493
TOK/S · 5% VRAM
›
S
Snowflake Arctic Embed M v2.00.305B
EMBEDDING·8K CTX· EMBEDDING· MULTILINGUAL
1322
TOK/S · 6% VRAM
›
S
bge-large-en-v1.5 335M0.335B
EMBEDDING·1K CTX· CHAT
1204
TOK/S · 6% VRAM
›
S
mxbai-embed-large-v10.335B
EMBEDDING·1K CTX· EMBEDDING
1204
TOK/S · 6% VRAM
›
S
Snowflake Arctic Embed L0.335B
EMBEDDING·1K CTX· EMBEDDING
1204
TOK/S · 6% VRAM
›
S
LFM2 350M0.35B
LFM·125K CTX· CHAT· TOOL_USE
1152
TOK/S · 6% VRAM
›
S
SmolLM2 360M0.36B
SMOLLM·8K CTX· CHAT
1120
TOK/S · 6% VRAM
›
S
GPT-2 Medium 345M0.38B
GPT2·1K CTX· CHAT
1061
TOK/S · 6% VRAM
›
S
Qwen 2.5 0.5B0.5B
QWEN·32K CTX· CHAT
806
TOK/S · 7% VRAM
›
S
SmolVLM 500M0.5B
SMOLLM·8K CTX· CHAT· VISION
806
TOK/S · 7% VRAM
›
S
Falcon-H1 0.5B0.52B
FALCON·128K CTX· CHAT
775
TOK/S · 7% VRAM
›
S
BGE-M30.568B
EMBEDDING·8K CTX· EMBEDDING
710
TOK/S · 7% VRAM
›
S
Snowflake Arctic Embed L v2.00.568B
EMBEDDING·8K CTX· EMBEDDING· MULTILINGUAL
710
TOK/S · 7% VRAM
›
S
Qwen3-Embedding 0.6B0.6B
EMBEDDING·32K CTX· EMBEDDING· MULTILINGUAL
672
TOK/S · 7% VRAM
›
S
Falcon-H1R Tiny 0.6B0.6B
FALCON·32K CTX· CHAT· REASONING
672
TOK/S · 7% VRAM
›
S
Falcon Perception 0.6B0.6B
FALCON·4K CTX· VISION
672
TOK/S · 7% VRAM
›
S
Qwen 1.5 0.5B0.62B
QWEN·32K CTX· CHAT
650
TOK/S · 7% VRAM
›
S
LFM2 700M0.74B
LFM·125K CTX· CHAT· TOOL_USE
545
TOK/S · 8% VRAM
›
S
Qwen3 0.6B0.75B
QWEN·32K CTX· CHAT· REASONING
538
TOK/S · 8% VRAM
›
S
GPT-2 Large 774M0.81B
GPT2·1K CTX· CHAT
498
TOK/S · 8% VRAM
›
S
Qwen 3.5 0.8B0.87B
QWEN·256K CTX· CHAT· CODING· MULTILINGUAL
463
TOK/S · 9% VRAM
›
S
InternVL3 1B0.94B
OTHER·32K CTX· CHAT· VISION
429
TOK/S · 9% VRAM
›
S
MiniCPM5 1B1.08B
MINICPM·128K CTX· CHAT· REASONING· MULTILINGUAL
373
TOK/S · 10% VRAM
›
S
TinyLlama 1.1B1.1B
LLAMA·2K CTX· CHAT
367
TOK/S · 10% VRAM
›
S
LFM2.5-1.2B-Thinking1.2B
LFM·122K CTX· CHAT· REASONING· TOOL_USE
336
TOK/S · 10% VRAM
›
S
Llama-3.2-1B1.2B
LLAMA·4K CTX· CHAT
336
TOK/S · 10% VRAM
›
S
LFM2 1.2B1.2B
LFM·125K CTX· CHAT· TOOL_USE· MULTILINGUAL
336
TOK/S · 10% VRAM
›
S
Zamba2 1.2B1.2B
OTHER·4K CTX· CHAT
336
TOK/S · 10% VRAM
›
S
EXAONE-4.0-1.2B1.3B
EXAONE·64K CTX· CHAT
310
TOK/S · 11% VRAM
›
S
OPT 1.3B1.3B
OPT·2K CTX· CHAT
310
TOK/S · 11% VRAM
›
S
MiniCPM-V 4.61.3B
OTHER·256K CTX· CHAT· VISION
310
TOK/S · 11% VRAM
›
S
DeepSeek Coder 1.3B1.35B
DEEPSEEK·16K CTX· CODING
299
TOK/S · 11% VRAM
›
S
Phi-1 1.3B1.42B
PHI·2K CTX· CODING
284
TOK/S · 11% VRAM
›
S
Phi-1.5 1.3B1.42B
PHI·2K CTX· CHAT· CODING
284
TOK/S · 11% VRAM
›
S
LFM2 8B A1B8.3BMoE
LFM·125K CTX· CHAT· TOOL_USE· MULTILINGUAL
269
TOK/S · 46% VRAM
›
▸ NEXT STEP

Get personalized recommendations.

See ranked models with benchmark scores, run commands, and precise speed estimates for your GeForce RTX 4070 Ti.

▸Find models for GeForce RTX 4070 TiCompare with other GPUs
WHAT THIS CARD IS WORTH

RTX 4070 Ti holds 248 of the models in our catalogue and is, in practice, a IQ4_XS card — the largest it takes is Ling-lite 16.8B at IQ4_XS.

TOKENS/SEC PER $100
6.9
8B at Q4_K_M, so cards compare like for like
VRAM PER $100
1.50 GB
what memory costs on this card
THE BIGGEST IT TAKES
Ling-lite 16.8B16.8B · IQ4_XS10.7 GB183 tok/sEST
DeepSeek V2 Lite 16B16B · Q4_K_S10.6 GB175 tok/sEST
StarCoder2 15B15.96B · Q4_K_S10.7 GB26 tok/sEST
DeepSeek-Coder-V2-Lite 15.7B15.7B · Q4_K_S10.4 GB175 tok/sEST
DeepSeek R1 Distill Qwen 14B14.8B · Q4_K_S10.4 GB28 tok/sEST
DeepCoder 14B14.8B · Q4_K_S10.4 GB28 tok/sEST
Qwen2.5-Coder-14B14.8B · Q4_K_S10.4 GB28 tok/sEST
Qwen2.5-14B14.8B · Q4_K_S10.4 GB28 tok/sEST
AGAINST CARDS OF SIMILAR MEMORY
M4 Pro (24GB)16 GB$1,3992.5 tok/s per $100
M4 (24GB)16 GB$6992.9 tok/s per $100
M2 (24GB)16 GB$9991.7 tok/s per $100
M3 (24GB)16 GB$9991.7 tok/s per $100

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

NVIDIA GeForce RTX 4070 Ti — 12 GB VRAM.

▸ GEFORCE RTX 4070 TI SPEC
BRAND
NVIDIA
VRAM
12 GB GDDR6X
BANDWIDTH
504 GB/s
FP16 COMPUTE
40.1 TFLOPS
FP32 COMPUTE
40.1 TFLOPS
CUDA CORES
7,680
TENSOR CORES
240
TDP
285 W
ARCHITECTURE
Ada Lovelace
MSRP
$799
▸ AI CAPABILITY
248/ 449 models @ Q4

With 12 GB VRAM and 504 GB/s bandwidth, this GPU handles models up to 14.8B parameters.

Speed ≈ bandwidth / model_size × efficiency. A 7B model at Q4 runs at ~58 tok/s.

RENT IT — LIVE PRICES
Checking what the clouds are charging…
§ 01TOP MODELS FOR GEFORCE RTX 4070 TI
248 FIT · SHOWING 20
MODELSIZEVRAM Q4TOK/SAVG
DeepSeek R1 Distill Qwen 14B14.8B9.5 GB2743.9
DeepCoder 14B14.8B9.5 GB2738.7
Qwen2.5-Coder-14B14.8B9.5 GB2741.3
Qwen2.5-14B14.8B9.5 GB2741.3
Qwen3 14B14.8B9.5 GB2745.7
phi-4 14B14.66B9.4 GB2833.7
Phi-4-reasoning 14B14.66B9.4 GB2833.7
Phi-4-reasoning-plus 14B14.66B9.4 GB2875.5
Phi-3-medium-14b14B9.0 GB2933.7
Qwen 1.5 14B14B9.0 GB2941.3
Ministral 3 14B Reasoning13.95B9.0 GB29—
Baichuan2 13B13B8.4 GB3123.6
Llama 2 13B13B8.4 GB3117.2
CodeLlama 13B13B8.4 GB3119.7
Vicuna 13B13B8.4 GB3111.8
LLaMA 1 13B13B8.4 GB3132.9
OPT 13B13B8.4 GB3135.8
Orca 2 13B13B8.4 GB3125.4
WizardCoder Python 13B13B8.4 GB3160.1
WizardLM 13B13B8.4 GB3119.5