Is the RTX 4090 Worth $3,400 for Local LLMs? (2026)
The RTX 4090 launched at $1,599 in October 2022, and NVIDIA stopped making it in late 2024. In November 2026 a used one averages $3,426 on eBay US, up from about $2,600 in September. In September it cost more per GB/s of bandwidth than a new RTX 5090. That comparison flipped when the 5090 rose to about $6,500. This page gives the verdict at today's price for each type of buyer, shows what the same money buys instead, and works out how many rental hours the card has to replace.
Bottom Line
A used RTX 4090 costs about $3,400 as of November 2026 (about $2,600 in September). That is 2.1x its $1,599 launch price from 2022. Whether it is worth that depends on who you are.
| Buyer | Worth ~$3,400? | Why |
|---|---|---|
| You run long-context agents or coding on 24GB, need CUDA, and cannot spend about $6,500 | Yes | It processes prompts 2.3-2.7x faster than a 3090 in community llama.cpp tests. Long prompts wait less. |
| You already own one | Yes, keep it | It resells above its launch price. Nothing new under $4,000 is faster on CUDA. |
| You chat with 20-35B models | No | A used RTX 3090 ($1,607) has the same 24GB and generates only about 15% slower, for 47% of the price. |
| You want 70B dense or 100B+ MoE models | No | 24GB cannot hold Llama 3.3 70B Q4_K_M (42.5GB). Two used 3090s ($3,214) give 48GB for less money. Two R9700s give 64GB for $2,598 at MSRP. |
| You can stretch to about $6,500 | Only for 32GB or top speed | The RTX 5090 adds 8GB and 1.78x the bandwidth for 1.9x the price. Per GB/s it now costs more ($3.63 against $3.40). |
| You use local AI a few hours a week | No | RunPod rents a 4090 for $0.34-0.89 per hour. The card needs about 4,230 hours of use to beat the higher rate. |
The short version: in November 2026 the 4090 costs twice as much per GB/s as a used 3090 ($3.40 against $1.72). It no longer costs more per GB/s than a new RTX 5090, because the 5090 rose to about $6,500 ($3.63 per GB/s). It sits between those two cards on speed. Buy it only for prefill speed on 24GB, at a budget that stops short of a 5090.
The price as of November 2026
NVIDIA told board partners to stop RTX 4090 production by October 2024 (TweakTown, 2024-10-06, citing Board Channels). New stock is now leftover inventory. The November 2026 prices:
| Measure | Price | Date | Source |
|---|---|---|---|
| Launch MSRP (Founders Edition) | $1,599 | 12 Oct 2022 | Wikipedia, GeForce RTX 40 series |
| Used, eBay US average | $3,426 | 2026-11-10 | ResalePrices (52 listings) |
| Used, eBay US median | $3,399 | 2026-11-10 | ResalePrices |
| Used, fair asking range | $3,255-3,544 | 2026-11-10 | ResalePrices |
| Used, 90-day low and high | $2,500-3,520 | to 2026-11-10 | ResalePrices (+14.2% in 30 days, +37.0% in 90 days) |
The September 2026 prices, kept as history:
| Measure | Price | Date | Source |
|---|---|---|---|
| Used, latest eBay sold reading | $2,510 | 2026-09-19 | GetPCParts |
| Used, 30-day average of eBay sold readings | $2,589 | to 2026-09-19 | GetPCParts (5 readings) |
| Used, highest reading this year | $3,010 | 2026-09-12 | GetPCParts |
| Used, eBay US median | $3,000 | 2026-09-25 | ResalePrices (17 listings, 90-day range $2,497-3,075, +20% in 30 days) |
| New, Newegg | $4,499 | read 2026-09-25 | GPUDojo |
| August 2026 used range (our earlier read) | $2,270-2,600 | August 2026 | ResalePrices / BestValueGPU |
We use $3,426 (the ResalePrices average) as the working price on this page. In September we used $2,589 (the GetPCParts 30-day sold average), because ResalePrices then had only 17 listings. The source changed, so read the two numbers as a direction, not an exact rise. In September GPUDojo also listed an eBay “new” price of $2,073. We treated that as one marketplace listing, not a market price.
The 4090 sells above its launch price, and the gap is growing. At $3,426 it costs 214% of its $1,599 MSRP (162% in September). A used RTX 3090 sells at 107% of its launch price. The cause is the 2026 memory shortage, plus a card that no one makes any more. Prices can also fall. A card at 2.1x its own MSRP has room to drop if new-card prices ease.
If you buy, compare the listing against the $3,255-3,544 fair range first. The catalog pick is the ASUS ROG Strix RTX 4090 24GB. New stock cost $4,499 at Newegg in September (GPUDojo, read 2026-09-25).
What $3,400 buys instead
Prices as of November 2026, with the read date in each row. Dollars per GB and per GB/s are our arithmetic, with the same method as our RTX 5090 and RTX 3090 verdicts. Bandwidth comes from vendor spec sheets.
| Option | Price | Memory | Bandwidth | $ per GB | $ per GB/s | Holds Llama 3.3 70B Q4_K_M (42.5GB)? |
|---|---|---|---|---|---|---|
| Used RTX 4090 | $3,426 avg (2026-11-10) | 24GB GDDR6X | 1,008 GB/s | $143 | $3.40 | No |
| Used RTX 3090 | $1,607 avg (2026-11-10) | 24GB GDDR6X | 936 GB/s | $67 | $1.72 | No |
| 2x used RTX 3090 | $3,214 | 48GB | 936 GB/s each | $67 | n/a | Yes |
| Radeon AI PRO R9700 | $1,299 MSRP; $1,700 3-month median; $2,630 lowest listing (GPUPrix, 2026-11-09) | 32GB GDDR6 | 640 GB/s | $41; $53; $82 | $2.03; $2.66; $4.11 | No |
| 2x R9700 | $2,598 at MSRP; $3,400 at the 3-month median | 64GB | 640 GB/s each | $41; $53 | n/a | Yes |
| RTX 5060 Ti 16GB (new) | $800 (GPUPrix, 2026-11-09) | 16GB GDDR7 | 448 GB/s | $50 | $1.79 | No |
| RTX 5090 (new) | about $6,500 (GPU Poet lowest average $6,331-6,643, 2026-11-08) | 32GB GDDR7 | 1,792 GB/s | $203 | $3.63 | No |
| RunPod RTX 4090 rental | $0.34-0.89/hr | 24GB | 1,008 GB/s | n/a | n/a | No |
In September 2026 the used 4090 cost $2,589 ($2.57 per GB/s), the used 3090 $1,463 ($1.56) and the 5090 $4,299 ($2.40).
Three things stand out.
- On capacity, the 4090 is the second-worst buy. At $143 per GB it costs 2.1x a used 3090 per GB. Only the 5090 costs more, at $203.
- On bandwidth, the 4090 costs twice the 3090, but it is no longer the worst buy. At $3.40 per GB/s it costs more than the 3090 ($1.72) and less than the 5090 ($3.63). In September it cost more than the 5090 ($2.57 against $2.40). Token generation speed follows bandwidth.
- The price of one used 4090 buys two cards with more memory. Two used 3090s cost $3,214 and give 48GB. Two R9700s cost $3,400 at the 3-month median price and give 64GB. Each pair holds a 70B model at Q4_K_M. The R9700 price is the catch: the lowest US listing on 2026-11-09 was $2,630 for one card.
The three alternatives, in numbers
Used RTX 3090. Same 24GB, same model fit. Its 936 GB/s is 93% of the 4090’s 1,008 GB/s. On the llama.cpp CUDA scoreboard it generated 158-162 tok/s on Llama 2 7B Q4_0, against 186-189 for the 4090. ResalePrices shows a $1,607 average across 326 eBay US listings on 2026-11-10 ($1,463 in September). A used RTX 3090 24GB (Amazon Renewed) costs 47% of a 4090. Pick a model with which used RTX 3090 to buy.
RTX 5090. It costs 1.9x the used 4090 (about $6,500 against $3,426). It has 1.78x the bandwidth and 8GB more memory. In community tests it generated 1.3-1.56x faster (details below). NVIDIA still sells new 5090s with a warranty. Used 4090s have none from the seller. In September the 5090 cost $4,299 at best, and it was the better use of money. At about $6,500, buy the RTX 5090 only if you need 32GB or the highest speed on one card. The verdict is in is the RTX 5090 worth it.
Radeon AI PRO R9700 32GB. It has 8GB more than the 4090. Its price moves a lot: $1,299 MSRP, a $1,700 3-month median, and $2,630 for the lowest US listing on 2026-11-09 (GPUPrix). That is 38-77% of a used 4090. It has 63% of the 4090’s bandwidth (640 against 1,008 GB/s), so expect slower generation on a model that fits both. That is a bandwidth-ratio estimate, not a measurement. It runs on ROCm, not CUDA. Buy the Radeon AI PRO R9700 32GB near its $1,299 MSRP. At $2,630 it costs more per GB/s than the 4090 ($4.11 against $3.40). The comparison against the 3090 is in R9700 vs RTX 3090.
RTX 4090 LLM performance on models that fit in 24GB
All numbers below are community-reported llama-bench results from the llama.cpp GitHub discussions. They are not our measurements.
| Model | Test | RTX 3090 | RTX 4090 | RTX 5090 | Source |
|---|---|---|---|---|---|
| Llama 2 7B Q4_0 | Generation (tg128), FA off | 158.2 tok/s | 186.2 tok/s | 290.0 tok/s | Discussion #15013 |
| Llama 2 7B Q4_0 | Prompt (pp512), FA off | 5,175 tok/s | 11,993 tok/s | 14,073 tok/s | Discussion #15013 |
| Llama 2 7B Q4_0 | Generation, FA on | 161.9 tok/s | 189.0 tok/s | not listed | Discussion #15013 |
| Llama 2 7B Q4_0 | Prompt, FA on | 5,560 tok/s | 14,771 tok/s | not listed | Discussion #15013 |
| Qwen3 30B-A3B Q4_K | Generation, CUDA graph opt on | not tested | 271.0 tok/s | 352.1 tok/s | Discussion #17621 (Nov 2025) |
| gpt-oss 20B MXFP4 | Generation, CUDA graph opt on | not tested | 272.0 tok/s | 419.1 tok/s | Discussion #17621 (Nov 2025) |
The 3090 and 4090 rows on the scoreboard come from different llama.cpp builds. Read them as a guide, not an exact ratio.
What the numbers say:
- Generation: the 4090 is only 17-18% faster than a 3090. 186.2 / 158.2 = 1.18. 189.0 / 161.9 = 1.17. The bandwidth gap is 8% (1,008 / 936).
- Prompt processing: the 4090 is 2.3-2.7x faster than a 3090. 11,993 / 5,175 = 2.3. 14,771 / 5,560 = 2.7. This is where the extra money goes.
- The 5090 is 1.3-1.56x faster than the 4090 at generation. 290.0 / 186.2 = 1.56. 352.1 / 271.0 = 1.30. 419.1 / 272.0 = 1.54.
Now divide the price by the Llama 2 7B generation speed (FA off). Our arithmetic:
- Used RTX 3090: $1,607 / 158.2 = $10.16 per tok/s
- Used RTX 4090: $3,426 / 186.2 = $18.40 per tok/s
- RTX 5090: $6,500 / 290.0 = $22.41 per tok/s
The 4090 costs 81% more per unit of generation speed than the 3090. It costs 18% less than the 5090.
What 24GB runs
File sizes from the Hugging Face API, as listed on our RTX 3090 verdict (read 2026-09-24).
| Model | Quant | File size | Fits 24GB? | 4090 generation |
|---|---|---|---|---|
| gpt-oss 20B | native MXFP4 | 13.8GB | Yes | 272 tok/s (community, #17621) |
| Qwen3.6 27B | Q4_K_M | 16.8GB | Yes | ~48 tok/s (our estimate, below) |
| Qwen3.6 35B-A3B | MXFP4_MOE | 21.7GB | Yes, with little context room | not measured |
| Llama 3.3 70B | Q4_K_M | 42.5GB | No | n/a |
| gpt-oss 120B | MXFP4 | 65.4GB | No | n/a |
The Qwen3.6 27B figure is an estimate. A dense model reads its full weights for each token. 1,008 GB/s / 16.8GB = 60 tok/s as a ceiling. The #17621 author reached about 80% of theoretical bandwidth on a 5090. 60 x 0.8 = 48 tok/s. The model picks for this card are in best local LLM for the RTX 4090. For 70B, read can an RTX 4090 run a 70B model.
Rent vs buy
RunPod lists the RTX 4090 at $0.89 per hour on Secure Cloud and $0.34 on Community Cloud (runpod.io/pricing, updated 2026-11-04, read 2026-11-10). In September the Secure rate was $0.74. Owning the card also costs power. NVIDIA rates the 4090 at 450W total graphics power. At $0.18/kWh (US average, as of August 2026), that is about $0.08 per hour under load.
Our arithmetic, for $3,426 and the November rates:
| Rental rate | Net saving per hour of ownership | Break-even hours | At 4 h/day | At 8 h/day | 24/7 |
|---|---|---|---|---|---|
| $0.89 (Secure) | $0.81 | ~4,230 | 2.9 years | 1.5 years | 5.8 months |
| $0.34 (Community) | $0.26 | ~13,230 | 9.1 years | 4.5 years | 18.1 months |
The formula is $3,426 / ($0.89 - $0.081) = 4,235 hours. At the Community rate it is $3,426 / ($0.34 - $0.081) = 13,228 hours. In September, at $2,589 and the $0.74 Secure rate, the break-even was 3,929 hours.
Two things favor buying. A card you own keeps resale value, and today’s resale value is above launch price. Your prompts also never leave your machine. One thing favors renting: the 4090 is nearly four years old and has no seller warranty. If you cannot name the job that fills 4 hours a day, rent for a month and log your hours first.
Is the RTX 4090 worth it over a used RTX 3090?
For chat, no. Both cards hold the same 24GB and run the same models. The 4090 generates 17-18% faster and costs 2.1x as much ($3,426 / $1,607).
For agents and coding, sometimes. Agent loops resend 20-30K tokens of context on every call. The 4090 processes prompts 2.3-2.7x faster than a 3090 on the scoreboard test. If your work waits on prompt processing, the 4090 pays for that wait. The full head-to-head is in RTX 3090 vs RTX 4090.
Is the RTX 4090 worth it over an RTX 5090?
On price per GB/s, yes, by a small margin. The 5090 costs 1.9x the used 4090 (about $6,500 against $3,426). It gives 1.78x the bandwidth, 1.33x the memory (32GB against 24GB), and a new-card warranty. Per GB/s the 4090 is now 6% cheaper ($3.40 against $3.63). In September the 5090 was the cheaper card per GB/s. Buy the 5090 if you need 32GB or the highest speed on one card. The three-way comparison is in RTX 5090 vs 4090 vs used 3090.
Power and fit
NVIDIA rates the 4090 at 450W and specifies an 850W minimum system power supply. A used card has no warranty from the seller. Stress-test it before the return window closes. Our used RTX 3090 checklist applies to the 4090 as well.
Sources
- GetPCParts: used RTX 4090 eBay sold prices, read 2026-09-25 (latest reading 2026-09-19)
- ResalePrices: used RTX 4090 eBay US market, updated 2026-11-10 (September read: 2026-09-25)
- GPUDojo: RTX 4090 new prices by retailer, read 2026-09-25
- TweakTown: NVIDIA to cease RTX 4090 production by October 2024, 2024-10-06
- NVIDIA GeForce RTX 4090 specifications (24GB GDDR6X, 384-bit, 450W, 850W system power)
- Wikipedia: GeForce RTX 40 series (launch 12 Oct 2022, $1,599 MSRP, 1,008 GB/s)
- llama.cpp discussion #15013: performance of llama.cpp on NVIDIA CUDA (community-reported)
- llama.cpp discussion #17621: optimizing token generation in the CUDA backend, 2025-11-30 (community-reported)
- ResalePrices: used RTX 3090 eBay US market, updated 2026-11-10 (September read: 2026-09-24)
- GPU Poet: RTX 5090 prices, read 2026-11-07 and 2026-11-08
- GPUPrix: Radeon AI PRO R9700 US price, updated 2026-11-09
- Tom’s Hardware: MSI RTX 5090 for $4,299 at Walmart, 2026-09-22
- AMD Radeon AI PRO R9700 specifications (32GB, 640 GB/s)
- runaihome: R9700 street price tracking, September 2026 (history)
- GPUPrix: RTX 5060 Ti 16GB US price, updated 2026-11-09
- RunPod GPU pricing, updated 2026-11-04, read 2026-11-10 (September read: 2026-09-22)
- Model file sizes from the Hugging Face API: unsloth/gpt-oss-20b-GGUF, unsloth/Qwen3.6-27B-GGUF, unsloth/Qwen3.6-35B-A3B-GGUF, bartowski/Llama-3.3-70B-Instruct-GGUF, unsloth/gpt-oss-120b-GGUF
See Also
- Best Local LLM for the RTX 4090: what to run on 24GB once you own one
- Is the RTX 5090 Worth It?: the 32GB flagship at about $6,500
- Is a Used RTX 3090 Worth It?: the same 24GB at about $1,600
- RTX 5090 vs 4090 vs Used 3090: the three-card buying rule
- RTX 3090 vs RTX 4090 for Local LLMs: same VRAM, different speed
- Which Used RTX 3090 Should You Buy?: pick on cooling and slot width
- R9700 vs RTX 3090 for Local LLMs: 32GB AMD against 24GB CUDA
- Can an RTX 4090 Run a 70B Model?: the limit of one card