NVIDIA · Ada Lovelace · 16 GB VRAM tier
Best AI Models for NVIDIA GeForce RTX 4080 SUPER (16 GB)
Runs 14B models well; 27–31B only just fits.
| # | Model | Speed | llmrun Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Qwen3.5 9BAlibabaHow to run → Q4_K_M · 6.6 GB · 75.2 t/s | 75.2 tok/s | Q4_K_M · 6.6 GB | |
| 2 | Qwen3.5 4BAlibaba Q4_K_M · 3.4 GB · 147 t/s | 147 tok/s | Q4_K_M · 3.4 GB | |
| 3 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB · 149 t/s | 149 tok/s | Q4_K_M · 13.1 GB | |
| 4 | Phi 4Microsoft Q4_K_M · 12.5 GB · 50.3 t/s | 50.3 tok/s | Q4_K_M · 12.5 GB | |
| 5 | Gemma 4 12B ITGoogle Q4_K_M · 9.0 GB · 58.1 t/s | 58.1 tok/s | Q4_K_M · 9.0 GB |
| # | Model | Speed | Coding Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | GPT OSS 20BOpenAIHow to run → Q4_K_M · 13.1 GB · 149 t/s | 149 tok/s | Q4_K_M · 13.1 GB | |
| 2 | Q4_K_M · 7.9 GB · 90.3 t/s | 90.3 tok/s | Q4_K_M · 7.9 GB | |
| 3 | Qwen3.5 9BAlibaba Q4_K_M · 6.6 GB · 75.2 t/s | 75.2 tok/s | overall | Q4_K_M · 6.6 GB |
| 4 | Gemma 4 12B ITGoogle Q4_K_M · 9.0 GB · 58.1 t/s | 58.1 tok/s | Q4_K_M · 9.0 GB | |
| 5 | Ling 3.0 TinyInclusion AI Q4_K_M · 5.5 GB · 305 t/s | 305 tok/s | Q4_K_M · 5.5 GB |
| # | Model | Speed | Reasoning Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Qwen3.5 9BAlibabaHow to run → Q4_K_M · 6.6 GB · 75.2 t/s | 75.2 tok/s | Q4_K_M · 6.6 GB | |
| 2 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB · 149 t/s | 149 tok/s | Q4_K_M · 13.1 GB | |
| 3 | Qwen3 8BAlibaba Q4_K_M · 7.6 GB · 86.7 t/s | 86.7 tok/s | Q4_K_M · 7.6 GB | |
| 4 | Q4_K_M · 7.9 GB · 90.3 t/s | 90.3 tok/s | Q4_K_M · 7.9 GB | |
| 5 | Gemma 4 12B ITGoogle Q4_K_M · 9.0 GB · 58.1 t/s | 58.1 tok/s | Q4_K_M · 9.0 GB |
| # | Model | Speed | Long context Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Qwen3.5 9BAlibabaHow to run → Q4_K_M · 6.6 GB · 75.2 t/s | 75.2 tok/s | Q4_K_M · 6.6 GB | |
| 2 | Qwen3.5 4BAlibaba Q4_K_M · 3.4 GB · 147 t/s | 147 tok/s | Q4_K_M · 3.4 GB | |
| 3 | Mistral Nemo Instruct 2407Mistral AI Q4_K_M · 11.0 GB · 59.3 t/s | 59.3 tok/s | Q4_K_M · 11.0 GB | |
| 4 | Q4_K_M · 7.9 GB · 90.3 t/s | 90.3 tok/s | Q4_K_M · 7.9 GB | |
| 5 | Gemma 4 12B ITGoogle Q4_K_M · 9.0 GB · 58.1 t/s | 58.1 tok/s | Q4_K_M · 9.0 GB |
| # | Model | Speed ↓ | llmrun Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | GPT OSS 20BOpenAIHow to run → Q4_K_M · 13.1 GB · 149 t/s | 149 tok/s | Q4_K_M · 13.1 GB | |
| 2 | Qwen3.5 4BAlibaba Q4_K_M · 3.4 GB · 147 t/s | 147 tok/s | Q4_K_M · 3.4 GB | |
| 3 | Qwen3.5 9BAlibaba Q4_K_M · 6.6 GB · 75.2 t/s | 75.2 tok/s | Q4_K_M · 6.6 GB | |
| 4 | Gemma 4 12B ITGoogle Q4_K_M · 9.0 GB · 58.1 t/s | 58.1 tok/s | Q4_K_M · 9.0 GB | |
| 5 | Phi 4Microsoft Q4_K_M · 12.5 GB · 50.3 t/s | 50.3 tok/s | Q4_K_M · 12.5 GB |
| # | Model | Speed | Vision Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Qwen3.5 9BAlibabaHow to run → Q4_K_M · 6.6 GB · 75.2 t/s | 75.2 tok/s | Q4_K_M · 6.6 GB | |
| 2 | Qwen3.5 4BAlibaba Q4_K_M · 3.4 GB · 147 t/s | 147 tok/s | Q4_K_M · 3.4 GB | |
| 3 | Gemma 3 12B ITGoogle Q4_K_M · 10.9 GB · 59.5 t/s | 59.5 tok/s | Q4_K_M · 10.9 GB | |
| 4 | Gemma 3 4B ITGoogle Q4_K_M · 5.0 GB · 169 t/s | 169 tok/s | Q4_K_M · 5.0 GB | |
| 5 | Q4_K_M · 6.5 GB · 77.0 t/s | 77.0 tok/s | Q4_K_M · 6.5 GB |
The RTX 4080 SUPER is a solid mid-range card in the 16 GB class: it runs 14B models well. It is more efficient per watt than most 16 GB cards, and pricier per token than most of the cards compared at ≈$1,155 used. Ada Lovelace generation, released 2024.
RTX 4080 SUPER strengths and weaknesses
Strengths
- Runs 35B MoE models: Qwen3.6 35B A3B at 153 tok/s in 15.7 GB
- Reads long prompts fast: A 4K-token prompt on Qwen3.5 9B starts answering in 1.7 s
Weaknesses
- 70B models don't fit: Llama 3.3 70B Instruct needs 51.3 GB at Q4 vs 16 GB here
- Used prices above MSRP: ≈$1,155 used vs its $999 MSRP (+16%)
Speed per dollar and per watt
On Qwen3.5 9B (9B params) at Q4_K_M. New prices are compared with new, used with used. The last two columns are the speed you get per $100 of price and per 100 W of board power; higher is better. The RTX 4080 SUPER has no new price, so the New view shows it at its MSRP.
| Card | New price | tok/s | tok/s per $100 | tok/s per 100 W |
|---|---|---|---|---|
RTX 507012 GB | about $840 | 68.7 | 8.2 | 27.5 |
RX 7900 XTX24 GB | about $1,130 | 90.6 | 8.0 | 25.5 |
RTX 5070 Ti16 GBBest new value | about $1,170 | 91.6 | 7.8 | 30.5 |
RX 907016 GB | about $800 | 60.4 | 7.6 | 27.5 |
RTX 4080 SUPER16 GB | about $999 | 75.2 | 7.5 | 23.5 |
RX 9070 XT16 GB | about $815 | 60.4 | 7.4 | 19.9 |
RTX 508016 GB | about $1,690 | 98.1 | 5.8 | 27.3 |
RX 9060 XT 16GB16 GB | about $590 | 30.2 | 5.1 | 18.9 |
| Card | Used price | tok/s | tok/s per $100 | tok/s per 100 W |
|---|---|---|---|---|
RTX 3060 12GB12 GBBest used value | about $329 | 36.8 | 11.2 | 21.6 |
RX 907016 GB | about $650 | 60.4 | 9.3 | 27.5 |
RX 7900 XTX24 GB | about $1,000 | 90.6 | 9.1 | 25.5 |
RTX 5070 Ti16 GB | about $1,100 | 91.6 | 8.3 | 30.5 |
RTX 4070 Ti SUPER16 GB | about $900 | 68.7 | 7.6 | 24.1 |
RTX 408016 GB | about $1,100 | 73.3 | 6.7 | 22.9 |
RTX 4080 SUPER16 GB | about $1,155 | 75.2 | 6.5 | 23.5 |
RX 9060 XT 16GB16 GB | about $480 | 30.2 | 6.3 | 18.9 |
Cost per token: the RTX 4080 SUPER against 12–32 GB cards
On Qwen3.5 9B (9B params, Q4_K_M, 4K context). Speed, cost per token and power per token depend on the model.
- Card, used price
- $0.97
- Power
- $0.18
Renting costs as much as buying after ≈5,000 hours (≈3.4 years at 4 h a day).
Each dot is a card; hover or tap it for its numbers, click to open it. Cards without a price for the chosen basis are left out. How we calculate cost per token
Cards people weigh against the RTX 4080 SUPER, each with the one reason it is worth a look. Speeds on Qwen3.5 9B.
Output speed depends on memory bandwidth (736 GB/s). Grey rows need more than 16 GB.
- 9BQwen3.5 9B · Q4_K_M · 6.6 GB75.2tok/s
- 14BQwen3 14B · Q4_K_M · 11.9 GB50.4tok/s
- 31.9tok/s
- 153tok/s
- 70BLlama 3.3 70B Instruct · not at 4-bitneeds 51.3 GB
- 120B MoEGPT OSS 120B · not at 4-bitneeds 70.8 GB
Estimated speeds: Q4_K_M, 4K context, one user. How we estimate
All 1057 models on the RTX 4080 SUPER
Featured onlyBest quant that fits, ordered by llmrun Score. Fit grades how well it runs here, S best. How ranking works
| Model | Quant | VRAM | Speed | Context | Score | Fit |
|---|---|---|---|---|---|---|
LFM2.5 VL 450MVisionFunctions | Q4_K_M | 0.8 GB | 747.5 t/s | 128K | S100, Runs great | |
| Q4_K_M | 5.6 GB | 277.8 t/s | 128K | S100, Runs great | ||
Olmo 3 7B ThinkChat | Q4_K_M | 8.4 GB | 83.2 t/s | 66K | S88, Runs great | |
Soren 1 SmallChatReasoningCodeMath | BF16 | 4.3 GB | 115.0 t/s | 1049K | S94, Runs great | |
Llama 3.2 3B InstructChatFunctions | Q4_K_M | 4.1 GB | 225.7 t/s | 131K | S100, Runs great | |
| Q4_K_M | 0.0 GB | 47840.0 t/s | 1K | S100, Runs great | ||
| Q4_K_M | 0.0 GB | 47840.0 t/s | 1K | S100, Runs great | ||
GPT S 5MChat | BF16 | 0.3 GB | 1495.0 t/s | 1K | S100, Runs great | |
Gemma 4 12B Agentic Fable5 Composer2.5 v2 3.5x Tau2ChatFunctionsReasoning | Q4_K_M | 9.0 GB | 58.1 t/s | 262K | A83, Runs well | |
InternVL3 8BVision | Q4_K_M | 7.8 GB | 91.3 t/s | — | S90, Runs great | |
ClefVisionFunctions | IQ4_XS | 15.9 GB | 31.0 t/s | 262K | C42, Usable | |
MiniCPM V 4.7 35B A3BVisionFunctions | IQ3_S | 15.9 GB | 170.9 t/s | 262K | C44, Usable | |
Qwen3.5 2B Claude 4.6 Opus Reasoning DistilledChatReasoning | Q4_K_M | 1.9 GB | 270.3 t/s | 262K | S100, Runs great | |
SmolVLM 256M InstructVision | Q4_K_M | 0.6 GB | 956.8 t/s | 8K | S100, Runs great | |
CyberPal2.0 20BChatReasoning | Q4_K_M | 13.1 GB | 148.5 t/s | 131K | S94, Runs great |
Specs
- Architecture
- Ada Lovelace
- VRAM
- 16 GB GDDR6X
- Memory bandwidth
- 736.0 GB/s
- FP16 (dense)
- 104.40 TFLOPS
- CUDA cores
- 10,240
- Tensor cores
- 320
- Compute capability
- 8.9
- TDP
- 320 W
- Released
- 2024-01-31
- MSRP
- $999
RTX 4080 SUPER FAQ
Is the RTX 4080 SUPER still worth it in 2026?
At ≈$1,155 used you get 6.5 tok/s per $100 on Qwen3.5 9B, which ranks 7th of the 8 cards compared here; it is also the 3rd fastest out of 17 cards with 16 GB. It runs 14B models well; 27–31B only just fits.
What is the largest model the RTX 4080 SUPER can run?
Gemma 4 26B A4B IT (25.8B) is the largest at 4-bit or better: at Q4_K_S it uses 15.5 GB of 16 GB, and runs at about 142 tok/s. At 3-bit, a bigger model fits too: Gemma 2 27B IT (27.2B) at IQ3_M, 15.0 GB.
Can the RTX 4080 SUPER run Qwen3.5 9B?
Yes. Qwen3.5 9B fits at Q4_K_M, using 6.6 GB of 16 GB, and it is the best model that runs well on this card.
Can the RTX 4080 SUPER run 70B models?
No. Llama 3.3 70B Instruct needs 51.3 GB at Q4, more than the 16 GB on this card.
How many tokens per second does the RTX 4080 SUPER give?
About 75.2 tok/s on Qwen3.5 9B at Q4_K_M, with the first token after 1.7 s on a 4K-token prompt.
What power supply does the RTX 4080 SUPER need?
Plan a 650 W PSU (2× the 320 W TDP).
Is 16 GB enough for coding and agents?
Yes for coding: GPT OSS 20B, the best of the 2 coding models that fit, runs at about 149 tok/s. For agents, not yet: no agent-scored model fits in less than 96 GB.
Is renting cheaper than buying an RTX 4080 SUPER?
Rent from $0.23/hr. Renting costs as much as buying after ≈5,000 hours (≈3.4 years at 4 h a day).