VOOZH about

URL: https://www.hardware-corner.net/gpu-llm-benchmarks/rtx-4090/

⇱ RTX 4090 Local LLM Benchmarks, Context Scaling & Supported Models 2026 – Hardware Corner

Tier 1 Enthusiast

RTX 4090 LLM Performance

Local LLM Performance: 78.7 t/s average on 14B models at 16k context. Updated Benchmarks: March 2026.

Gen (14B 4-bit) 78.7 t/s
PP (14B 4-bit) 4,474 t/s
Max Model 34B
VRAM
24 GB GDDR6
Bandwidth 1,008 GB/s
Token Gen (14B @ 4k Ctx)

78.7T/s

Prompt Proc (14B @ 4k Ctx)

4,474T/s

Summary

While we found the NVIDIA RTX 4090 to be faster than the RTX 3090 in our benchmarks, the practical difference in daily use wasn't massive. Since both cards share the same 24GB VRAM limit, our testing showed identical model capacity, with each handling 34B parameter models at 4-bit quantization. Given the current pricing, we don’t believe the performance gain justifies the premium, and we still point toward a second-hand 3090 as the smarter buy for value-focused users.

Key Insights

Capable of running Qwen3 34B using GGUF (Q4_K) completely in VRAM.
Handles context splitting effectively up to 16k tokens on 34B model.
Handles gpt-oss 120B model in MXFP4 quantization at full128K context.
Supports Flash Attention 2, significantly boosting prompt processing speeds.

Current Price in US

$2,200

Avg. Market Value

Current Pricing

Hardware Specs
VRAM 24GB GDDR6
Capable of running 34B model
Bandwidth 1,008 GB/s
Architecture Ada Lovelace
Memory speed 21 Gbps
Memory bus 384 bit
TDP 450 W
Suggested PSU 850 W
Price/GB VRAM $91.67
Price/(t/s) with 14B @ 16k $27.95

Biggest LLMs You Can Run on This GPU

The models below represent the largest language models that fit fully in VRAM on this GPU using 4-bit quantization (GGUF). Benchmarks include token generation and prompt processing speeds measured at their maximum supported context length.

Qwen3 32B (Q4_K) Max 16k
Token Generation 34.4 t/s @ 16k context
Prompt Processing 1,707.4 t/s @ 16k context
Qwen3 30B A3B (Q4_K) Max 64k
Token Generation 68.2 t/s @ 64k context
Prompt Processing 1,502.5 t/s @ 64k context
gpt-oss 20B (MXFP4) Max 128k
Token Generation 70.3 t/s @ 128k context
Prompt Processing 1,114.7 t/s @ 128k context

Note: Context values are grouped into standard tiers (4K, 16K, 32K, 64K, 128K). Models may support slightly higher context, but they remain in the lower tier unless they reach the next bracket.

RTX 4090 local LLM Inference Performance vs Similar GPUs

Compare prompt ingestion and token generation speeds against similar GPUs across widely used local models and extended context lengths up to 256K.

Local LLM Benchmarks

Prompt processing (t/s) and token generation speed (t/s) across different open weight models and context lengths.

Prompt Processing
Model 4k Ctx 16k Ctx 32k Ctx 64k Ctx 128k Ctx 256k Ctx
Qwen3 8B (Q4_K) 9,250.5 5,530.5 3,560.1 2,028.6 1,059.9 —
Qwen3 14B (Q4_K) 5,265.4 3,452.3 2,324.2 1,398.3 — —
gpt-oss 20B (MXFP4) 8,369.3 5,856.2 3,705.6 2,092.2 1,114.7 —
Qwen3 30B A3B (Q4_K) 6,159.4 4,027.3 2,627.7 1,502.5 — —
Qwen3 32B (Q4_K) 2,448.0 1,707.4 — — — —
Token Generation
Model 4k Ctx 16k Ctx 32k Ctx 64k Ctx 128k Ctx 256k Ctx
Qwen3 8B (Q4_K) 141.3 108.0 82.3 56.1 33.8 —
Qwen3 14B (Q4_K) 84.4 69.8 55.4 38.8 — —
gpt-oss 20B (MXFP4) 190.6 163.3 140.7 104.0 70.3 —
Qwen3 30B A3B (Q4_K) 207.3 139.2 105.1 68.2 — —
Qwen3 32B (Q4_K) 39.6 34.4 — — — —

Frequently Asked Questions

Common questions about running LLMs on the RTX 4090.