GeForce RTX 4060 Ti 16 GB
Data checked: 2026-10-12. Review interval shown below. Prices need review after 30 days; specifications and benchmark collections after 90 days. Dates describe the recorded check, not a live retailer quote.
- Verified Checked against the primary source (for example the model’s config.json on Hugging Face).
- Sourced Taken from a cited source; not yet independently checked.
- Estimate Calculated with the published formulas; not a measurement.
- Sample Placeholder data for layout; not a real figure.
- Not yet measured No public, citable score found for this benchmark yet. We never fill the gap with a guess.
Verified metadata is not a measured performance result. Only cited runs matching model, device, quant and recorded context can replace speed estimates. Estimates assume one stream and F16 KV cache, with a ±25% uncertainty band. A published run applies only to its recorded runtime and conditions; missing settings are unknown.
- llama-2-7b · rtx-4060-ti-16g
llama.cpp (CUDA) · Q4_0 · tg128
Recorded 2026-10-12 · historical source record · repeat count unknownSourced
Llama 2 7B Q4_0, llama-bench tg128, full GPU offload, flash attention off, single GPU. Community scoreboard entry; commit and submitter in the source table. Measured on the 8 GB variant (same 288 GB/s bus); treat as indicative for the 16 GB card. Run source
Evidence labels · Freshness policy · Calculation methodology
Cheapest new 16 GB card. Narrow 128-bit bus caps decode speed.
Decode-speed ceiling for a model is 288 GB/s ÷ GB read per token. For a 16.6 GB file that is 17 tok/s; real runtimes reach roughly 60% of it. Why bandwidth is the limit.
In builds: Silent home server
See every model on this deviceSpecifications
Memory | 16 GBSourced16 GB variant, nvidia.com, checked 2026-10-12. Source |
|---|---|
Bandwidth | 288 GB/sSourcedNVIDIA's GeForce spec pages list bus width but not bandwidth; figure from the cited Wikipedia spec table (bus width cross-checked on nvidia.com, 2026-10-12). Source |
FP16 dense compute | 44 TFLOPSSampleSample FP16 dense compute; pending vendor-sheet citation. |
Host RAM (offload) | 32 GBSampleAssumed host RAM for offload. |
Vendor TDP / board power | 165 WSourcedTotal Graphics Power (16 GB variant), nvidia.com. Source |
Load power at wall | 230 WSampleSample wall power under load. |
Price, USD | $844.37EstimateNo valid new 16 GB listing found on Newegg US; Newegg Canada price ÷ 1.4271 (Bank of Canada). Source |
Price, CAD | CA$1,205SourcedLowest new in-stock 16 GB listing on Newegg Canada (MSI Gaming). Checked 2026-10-12. Source |
Models that run
Default quant, single stream, 8k context. Fastest first; estimates are ±25% until measured.
Model | Quant | Memory | Verdict | Decode, tok/s | Max context |
|---|---|---|---|---|---|
| gpt-oss-20b | MXFP4 | 12.9 GB | Fits | 75.5Estimate | 64k |
| Qwen3 30B-A3B | Q4_K_M | 19.9 GB | Needs offload | 35.1Estimate | — |
| Llama 3.1 8B Instruct | Q4_K_M | 6.6 GB | Fits | 28.8Estimate | 64k |
| Qwen3 8B | Q4_K_M | 6.8 GB | Fits | 27.7Estimate | 32k |
| Qwen3 14B | Q4_K_M | 10.9 GB | Fits | 16.7Estimate | 32k |
| Mistral Small 3.2 24B | Q4_K_M | 16.2 GB | Needs offload | 9.1Estimate | 2k |
| Gemma 3 27B IT | Q4_K_M | 18.2 GB | Needs offload | 5.8Estimate | — |
| Qwen3 32B | Q4_K_M | 22.5 GB | Needs offload | 3.3Estimate | — |
| Llama 3.3 70B Instruct | Q4_K_M | 45.8 GB | Does not fit | — | — |
| gpt-oss-120b | MXFP4 | 64.3 GB | Does not fit | — | — |
Published reference runs
Third-party llama.cpp measurements on this device, with conditions and source. Llama 2 7B Q4_0 runs are the community calibration test; they are not catalogue models.
Model | Quant | Date | Runtime · test | Decode, tok/s | Conditions |
|---|---|---|---|---|---|
| llama-2-7b | Q4_0 | 2026-10-12 | llama.cpp (CUDA) · tg128 | 63.9Sourced | Llama 2 7B Q4_0, llama-bench tg128, full GPU offload, flash attention off, single GPU. Community scoreboard entry; commit and submitter in the source table. Measured on the 8 GB variant (same 288 GB/s bus); treat as indicative for the 16 GB card. Source |