AMD Radeon RX 6700 XT — Benchmarks & Specs
Short answer: the AMD Radeon RX 6700 XT generates 16.4 tokens/sec on llama3.1:8b (q4_K_M) under ollama in a single cited run, per LocalScore; there are too few comparable runs on file for a median. Its 12 GB of VRAM holds up to Phi-4 14B at Q4_K_M with an ~8K context, by the Will-It-Run matrix's sizing (13 of 27 models fit).
- Single cited run 16.4 tok/s llama3.1:8b q4_K_M, per LocalScore
- Fits in 12 GB up to Phi-4 14B at Q4_K_M → at ~8K context; 13 of 27 models in the Will-It-Run matrix
- Lowest linked price $749.00 Amazon, checked ; prices change
Who should buy this instead: the AMD Radeon RX 6700 XT's 12 GB puts it in the 12 GB tier of the local-LLM GPU picks by VRAM (one card per tier, each with cited tok/s). If the models you run fit in 8 GB, compare the 8 GB pick. If they need more than 12 GB, start at the 16 GB pick.
AMD Radeon RX 6700 XT: verdict by use case
- For the largest model in 12 GB: Phi-4 14B at Q4_K_M with an ~8K context; 13 of the 27 models in the Will-It-Run matrix fit. By the same sizing, 65-73B models at Q4 do not fit, so 70B-class work needs a card with more VRAM or CPU offload.
- For gaming: 60 fps average in A Plague Tale: Legacy at 1440p Ultra, per PCBench.
Each verdict quotes a figure from a linked source or a row in the tables below. Where no cited figure exists for a use case, there is no verdict line for it.
Bottom line: how fast is the Radeon RX 6700 XT?
At 1440p Ultra, the Radeon RX 6700 XT averages 60 fps in A Plague Tale: Legacy with a 52 fps 1% low, per PCBench. For local LLM inference it generates 16.4 tokens/sec running llama3.1:8b at q4_K_M under ollama, per LocalScore. In PassMark G3D Mark it scores 19,730–19,732 points, across two PassMark results (PassMark GPU benchmark listing: 19,732 points, PassMark GPU benchmark listing: 19,730 points). It has 12 GB of VRAM, and the largest model on file at 4-bit quantization on this card is llama-2:7b (7B parameters, q4_0), generating 83.9 tokens/sec under llama.cpp, per llama.cpp GitHub Discussions. Larger models have no sourced run on file.
Every figure above is a row in the tables below, and each row links out to the review or public benchmark database the number was taken from. SpecPicks aggregates published measurements; it does not report first-party benchmark runs.
Lowest current Amazon price among the linked cards: $749.00, checked (price may vary)
This Amazon listing is above the Radeon RX 6700 XT's $479 launch MSRP. Compare new and used listings on eBay before buying.
*Price sourced from Amazon.com. Last updated 2026-10-26. Price and availability subject to change.
Read the full comparison → RTX 3060 12GB vs RX 6700 XT 12GB for Phi-4 14B (2026)
Check what fits in 12 GB: by file size at each listed quantization, 13 of the 27 open-weight model files in the SpecPicks Will-It-Run matrix load entirely into the AMD Radeon RX 6700 XT's 12 GB while keeping at least 1.5 GB reserved for an ~8K context; the largest that fits, Phi-4 14B (a 9.1 GB Q4_K_M file per its Hugging Face listing), leaves roughly 2.9 GB free. Check a model in the Will-It-Run matrix →
What fits in the AMD Radeon RX 6700 XT's 12 GB, and what doesn't
2 of these 7 model files load whole into 12 GB with room for an ~8K context. A file fits when the card keeps at least 1.5 GB free after the weights; file sizes are the GGUF downloads on Hugging Face. Speeds are this card's single-user generation speed from cited community runs at the Q4-class file listed; a row with no cited run says so rather than estimating one.
| Model | Quant | Fits in 12 GB? | Cited speed on this card |
|---|---|---|---|
| Llama 3.1 8B 4.9 GB file | Q4_K_M | Fits~7.1 GB left for context | 16.4 tok/s1 cited run, ollama, per LocalScore (Mozilla Builders) |
| Qwen3 14B 9 GB file | Q4_K_M | Fits~3 GB left for context | No public measurement on file |
| gpt-oss-20b 21B-A3.6B · 12.1 GB file | MXFP4 | Doesn't fit0.1 GB over; runs only with CPU offload | — (does not load into VRAM) |
| Qwen3 30B-A3B 18.6 GB file | Q4_K_M | Doesn't fit6.6 GB over; runs only with CPU offload | — (does not load into VRAM) |
| Qwen3 32B 19.8 GB file | Q4_K_M | Doesn't fit7.8 GB over; runs only with CPU offload | — (does not load into VRAM) |
| Gemma 3 27B 28.7 GB file | Q8_0 | Doesn't fit16.7 GB over; runs only with CPU offload | — (does not load into VRAM) |
| Llama 3.3 70B 42.5 GB file | Q4_K_M | Doesn't fit30.5 GB over; runs only with CPU offload | — (does not load into VRAM) |
The Radeon RX 6700 XT is a graphics card from the RDNA 2 family released in 2021 from AMD. Key on-paper specs include 12 GB of VRAM. It launched with a $479 MSRP, though street prices typically diverge meaningfully from launch pricing — see the linked product cards below for current listings. Data on this page draws on 8 linked listings, 14 synthetic benchmark results (top 10 shown), 8 community AI inference reports, 63 measured game frame-rate results (top 15 shown), compiled from KitGuru, The FPS Review, Wccftech, LocalScore, AlkTech, llama.cpp GitHub Discussions, PCBench, TechPowerUp, geerlingguy/ai-benchmarks, LocalLLaMA, Ollama GitHub PR #814, PassMark; each table row links to its source. Read this page when shopping the Radeon RX 6700 XT, comparing it against other graphics cards in your build, or sizing it for a specific workload (gaming at 1080p/1440p/4K, synthetic benchmark scores, or local LLM inference).
Gaming Performance (measured FPS)
Average and 1% low frame rates by game, resolution, and quality preset. Bars are scaled against the fastest result on this page.
| Game | Resolution | Settings | Relative | Avg FPS | 1% low | Source |
|---|---|---|---|---|---|---|
| Counter-Strike 2 | 1080p | Ultra | 190 fps | 165 fps | TechPowerUp 2021-01-25 | |
| Hitman 3 | 1080p | Ultra | 180 fps | — | KitGuru 2021-03-17 | |
| Hitman 3 | 1440p | Ultra | 135 fps | — | The FPS Review 2021-03-17 | |
| Forza Horizon 4 | 1080p | Ultra | 132 fps | — | AlkTech 2022-08-08 | |
| Hitman 3 | 1440p | Ultra | 129 fps | — | KitGuru 2021-03-17 | |
| F1 2020 | 1440p | Max | 124 fps | — | The FPS Review 2021-06-22 | |
| Gears 5 | 1080p | Ultra | 111 fps | — | KitGuru 2021-03-17 | |
| Control | 1080p | Ultra | 109 fps | — | KitGuru 2021-03-17 | |
| The Division 2 | 1080p | Ultra | 105 fps | — | KitGuru 2021-03-17 | |
| Gears 5 | 1440p | Ultra | 101 fps | — | KitGuru 2021-03-17 | |
| Far Cry 6 | 1080p | Ultra | 98 fps | — | AlkTech 2022-08-08 | |
| Horizon Zero Dawn | 1440p | Ultimate Quality | 93 fps | — | The FPS Review 2021-06-22 | |
| Horizon Zero Dawn | 1440p | Ultra | 93 fps | — | The FPS Review 2021-06-22 | |
| Assassin's Creed Valhalla | 1080p | Ultra High | 93 fps | — | KitGuru 2021-03-17 | |
| A Plague Tale: Legacy | 1440p | Ultra | 60 fps | 52 fps | PCBench 2025-07 |
AI Inference Performance
Tokens per second under each model + quantization. Higher = faster generation. Bars compare runs across the same model.
| Model | Quantization | Relative | Tokens/sec | VRAM used | Source |
|---|---|---|---|---|---|
| qwen3:8b | Q8_0 llama.cpp | 711.0 tok/s | — | LocalLLaMA 2026-03-27 | |
| llama2:7b | Q4_0 llama.cpp | 83.9 tok/s | — | llama.cpp GitHub Discussion #10879 2024-06 | |
| llama-2:7b | q4_0 llama.cpp | 83.9 tok/s | — | llama.cpp GitHub Discussions | |
| llama3.2:1b | q4_K_M ollama | 64.3 tok/s | — | LocalScore (Mozilla Builders) retrieved 2026-08-18 | |
| codellama | — ollama | 56.3 tok/s | — | Ollama GitHub PR #814 2023-11-11 | |
| llama3.2:3b | — ollama | 49.0 tok/s | — | geerlingguy/ai-benchmarks 2024-12 | |
| llama3.1:8b | q4_K_M ollama | 16.4 tok/s | — | LocalScore (Mozilla Builders) retrieved 2026-08-18 | |
| qwen2.5:14b | q4_K_M ollama | 15.7 tok/s | — | LocalScore (Mozilla Builders) retrieved 2026-08-18 |
Synthetic Benchmarks
Higher is better. Bars are scaled within each benchmark family (multi-thread, single-thread, etc.) so you can compare like-with-like at a glance.
| Benchmark | Relative | Score | Source |
|---|---|---|---|
| PassMark G3D Mark | 19,732 points | PassMark Software retrieved 2026-04-28 | |
| PassMark G3D Mark | 19,730 pts | PassMark retrieved 2026-04-20 | |
| 3DMark Fire Strike Extreme (Graphics score, reported rounded) | 17,400 points | Wccftech 2021-03-17 | |
| 3DMark Time Spy | 14,000 pts | TechPowerUp 2021-01-25 | |
| 3DMark Time Spy (Graphics score, reported rounded) | 12,000 points | Wccftech 2021-03-17 | |
| 3DMark Time Spy | 10,779 points | The FPS Review 2021-06-22 | |
| 3DMark Fire Strike Ultra (Graphics score, reported rounded) | 8,500 points | Wccftech 2021-03-17 | |
| 3DMark Port Royal | 5,839 points | The FPS Review 2021-06-22 | |
| 3DMark Port Royal (reported rounded) | 5,800 points | Wccftech 2021-03-17 | |
| 3DMark Time Spy Extreme (Graphics score, reported rounded) | 5,600 points | Wccftech 2021-03-17 |
Products Featuring the Radeon RX 6700 XT
Full Specifications
| VRAM | 12 GB |
|---|
Radeon RX 6700 XT — Frequently Asked Questions
What is the Radeon RX 6700 XT best used for?
When was the Radeon RX 6700 XT released, and what was its launch MSRP?
Where do the Radeon RX 6700 XT benchmark numbers come from?
Can the Radeon RX 6700 XT run local LLMs?
Where can I buy the Radeon RX 6700 XT?
Buying guides that rank the Radeon RX 6700 XT's class
This page is the raw performance data. The guides below turn it into a ranked pick for a specific build.
Editorial guides covering the Radeon RX 6700 XT
In-depth SpecPicks reviews, build guides, and head-to-heads referencing this graphics card.
- RTX 3060 12GB vs RX 6700 XT 12GB for Phi-4 14B (2026)
- 9800X3D + RTX 5070 Ti PC Build 2026: 1440p High-Refresh
- 4K Gaming PC Build 2026: Cost, Parts, FPS
- $1,500 Gaming PC Build 2026: Parts, Cost, 1440p FPS
- Can You Build an $800 Gaming PC in 2026? Parts + Cost
- RX 9060 XT 16GB vs 8GB: Which One to Buy (Benchmarks)
- $1,200 Gaming PC Build 2026: Parts, Cost, 1440p FPS
- 1440p Gaming PC Build 2026: Cost, Parts, FPS
- Browse all SpecPicks reviews →
More guides & deep dives from the SpecPicks archive
Browse all articles & guides →- Best Retro Handhelds in 2026 — From $35 to $500
- Best 1440p Gaming GPUs in 2026
- The Complete Voodoo5 5500 AGP Driver Guide (2026 Edition)
- Best Budget Gaming PC Build 2026 — ~$1,000 Parts List
- Emulation Hardware in 2026: FPGA, Software, and Cart-Reader Ecosystems
- RTX 4070 Super vs RX 7800 XT — Which to Buy in 2026
- How to Build a Windows 98 Retro PC in 2026
More reviews from the SpecPicks archive
Browse all reviews →- Gemma-4-Harmonia-31B Uncensored on RTX 3060 12GB: Quantization, VRAM, and tok/s
- GPU VRAM Guide: Running a 5GB TTS Model Locally
- Best Streaming Starter Kits in 2026: Mic, Light, Capture, and the One-Box Builds
- Modern PC Builds vs Period-Correct Retro Rigs: A 2026 Comparison
- Raspberry Pi 4 8GB vs Ryzen 5 5600G: Which Homelab Box Wins in 2026?
- Best PC and Emulation Controllers in 2026
- Local Image Generation AI in 2026: AMD Hardware Performance & Setup Guide
- Sound Blaster Audigy FX in a Period-Correct Win98 Gaming Build
- Govee vs DAYBETTER vs KSIPZE: Best LED Strip for Desk and Monitor Bias Lighting
- Best SATA SSD for Gaming and Everyday Upgrades in 2026
- Best Mouse and Mousepad for FPS in 2026: G502 Hero + SteelSeries QcK
- Best Budget Gaming CPU: Ryzen 5 5600G vs 5700X vs i7-9700K
- Best Parts for an Always-On Local LLM Server in 2026
- How to run Qwen 3 32B on AMD Radeon RX 7900 XTX
- Which GPU for Which Model: A Per-LLM VRAM Picker for Local Rigs (2026)
- Best Gaming Room Lighting in 2026: LED Strips and Ring Lights
- RTX 3060 Driver Crashes and Stuttering: Troubleshooting Guide (2026)
- SNES Classic vs Raspberry Pi 4 RetroPie: Which Retro Console Wins in 2026?
- How to Image and Preserve a Vintage IDE Hard Drive in 2026 (CompactFlash + USB Adapter Workflow)
- Best Controller for PC Emulation and Retro Console Crossover (2026)
- Best GPUs for Mixtral 8x22B Local Inference in 2026
- Samsung 870 EVO vs WD Blue SN550: Best Storage for Forza Horizon 6 Fast Loading
- Best Budget 1TB SATA SSD 2026: BX500 vs WD Blue vs SanDisk Ultra 3D vs Samsung 870 EVO
- xAI Open-Sources Grok-Build: Can You Run It Locally?
More buying guides from SpecPicks
Browse all buying guides →- Best GPUs for Running Local LLMs in 2026
- Best AM5 Motherboards for 2026
- Best NVMe External Enclosures for 2026
- Best Controllers for PC Gaming in 2026
- Best GPUs for 4K Gaming in 2026
- Best NVMe SSDs for Gaming in 2026
- Best External SSDs for Content Creators in 2026
- Best Gaming Monitors for 2026
- Best Graphics Cards for Gaming in 2026
- Best CPUs for Gaming in 2026
- Best PC Cases for Building in 2026
- Best 1440p 240Hz Gaming Monitors in 2026
- Best Retro Gaming Consoles & Handhelds for 2026
- Best CPUs for Content Creators in 2026
- Best GPU for Running 27B-32B Local LLMs in 2026
- Best CPU Coolers for 2026
- Best 4K Monitors for Content Creators in 2026
- Best Mechanical Keyboards for Gaming in 2026
- Best Gaming Mice for 2026
- Best Tools for Building and Repairing Retro PCs in 2026
- Best DDR5 RAM for Gaming PCs in 2026
Hardware benchmark data on SpecPicks
All benchmarks →- Radeon RX 6750 XT — benchmarks & specs
- Quadro RTX 6000 — benchmarks & specs
- Apple M2 Max 12 Core 3680 MHz — benchmarks & specs
- Apple M4 9 Core — benchmarks & specs
- Radeon RX 6800 XT — benchmarks & specs
- Ryzen 7 5800H — benchmarks & specs
- GeForce RTX 4060 Ti 16GB — benchmarks & specs
- AMD Ryzen Threadripper PRO 5955WX — benchmarks & specs
- AMD Ryzen Threadripper PRO 7985WX — benchmarks & specs
- NVIDIA GeForce RTX 4070 Ti SUPER — benchmarks & specs
- AMD Ryzen 5 5500GT — benchmarks & specs
- Ryzen 7 8845HS — benchmarks & specs
- AMD Ryzen 5 7645HX — benchmarks & specs
- NVIDIA GeForce GTX 1650 — benchmarks & specs
- NVIDIA GeForce RTX 5060 Ti — benchmarks & specs
- AMD Ryzen 5 PRO 5675U — benchmarks & specs
- AMD Ryzen 3 7320C — benchmarks & specs
- NVIDIA Jetson Orin NX 16GB — benchmarks & specs
- NVIDIA RTX A5500 24GB — benchmarks & specs
- AMD Ryzen 9 PRO 7945 — benchmarks & specs
- GRID RTX6000P-2B — benchmarks & specs
- NVIDIA RTX A4000 16GB — benchmarks & specs
- AMD Ryzen 9 7900X — benchmarks & specs
- Ryzen 9 5900X — benchmarks & specs