Compare · cited /models pages · llama.cpp
Compare two models
Pick two models that already have cited AMDai /models pages — Qwen3.6-35B, Qwen 3.8 Flash, gpt-oss 20B, Gemma 3 27B, Llama 3.3 70B, GLM 5.3 Flash, Qwen3.6 27B, and Qwen3.8 27B. Each pair page uses official-card sizes and context claims, calculator fit on the existing R9700 / RX 7900 XTX / RX 7900 XT / RX 7800 XT sheets at Q6 / Q5 / Q4 (no Q3), and llama.cpp-first commands already on those model pages. Measured decode appears only from existing local desktop run cards. We do not invent a price or a benchmark. Minimum system RAM needed and RAM needed at full context print wherever a file spills to system RAM.
How we test — Measured cells cite official amdai-bench/run@1 cards. Estimates stay labeled. Q3 is estimate-only. Q1/Q2 stay off.
Useful pairs
Pre-rendered, shareable URLs. Every other catalog pair is also a static page — use the picker or the lists below.
-
35B total / 3B active vs 320B total / 18B active
Qwen3.6-35B vs GLM 5.3 Flash
Qwen3.6-35B is 35B total / 3B active, 262,144 tokens on the official card. GLM 5.3 Flash is 320B total / 18B active, 1,048,576 tokens. Q6 / Q5 / Q4 file sizes come from the listed GGUFs on each /models page, or a calculator estimate when a listing is missing. Q3 stays off. Q4 fits Qwen3.6-35B on one R9700 at 4K and spills on GLM 5.3 Flash. Qwen3.6-35B has measured llama.cpp cells on the local desktop. GLM 5.3 Flash speed is not measured yet. llama.cpp first.
-
35B total / 3B active vs 125B total / 6B active
Qwen3.6-35B vs Qwen 3.8 Flash
Qwen3.6-35B is 35B total / 3B active, 262,144 tokens on the official card. Qwen 3.8 Flash is 125B total / 6B active, 262,144 tokens. Q6 / Q5 / Q4 file sizes come from the listed GGUFs on each /models page, or a calculator estimate when a listing is missing. Q3 stays off. Q4 fits Qwen3.6-35B on one R9700 at 4K and spills on Qwen 3.8 Flash. Qwen3.6-35B has measured llama.cpp cells on the local desktop. Qwen 3.8 Flash speed is not measured yet. llama.cpp first.
-
35B total / 3B active vs 27.4B
Qwen3.6-35B vs Gemma 3 27B
Qwen3.6-35B is 35B total / 3B active, 262,144 tokens on the official card. Gemma 3 27B is 27.4B, 128K tokens. Q6 / Q5 / Q4 file sizes come from the listed GGUFs on each /models page, or a calculator estimate when a listing is missing. Q3 stays off. Q4 files fit one R9700 at 4K on the calculator for both sides. Qwen3.6-35B has measured llama.cpp cells on the local desktop. Gemma 3 27B speed is not measured yet. llama.cpp first.
-
35B total / 3B active vs 70.6B
Qwen3.6-35B vs Llama 3.3 70B
Qwen3.6-35B is 35B total / 3B active, 262,144 tokens on the official card. Llama 3.3 70B is 70.6B, 128K tokens. Q6 / Q5 / Q4 file sizes come from the listed GGUFs on each /models page, or a calculator estimate when a listing is missing. Q3 stays off. Q4 fits Qwen3.6-35B on one R9700 at 4K and spills on Llama 3.3 70B. Qwen3.6-35B has measured llama.cpp cells on the local desktop. Llama 3.3 70B speed is not measured yet. llama.cpp first.
-
35B total / 3B active vs 21.5B total / 3.6B active
Qwen3.6-35B vs gpt-oss 20B
Qwen3.6-35B is 35B total / 3B active, 262,144 tokens on the official card. gpt-oss 20B is 21.5B total / 3.6B active, 131,072 tokens. Q6 / Q5 / Q4 file sizes come from the listed GGUFs on each /models page, or a calculator estimate when a listing is missing. Q3 stays off. Q4 files fit one R9700 at 4K on the calculator for both sides. Qwen3.6-35B has measured llama.cpp cells on the local desktop. gpt-oss 20B speed is not measured yet. llama.cpp first.
-
35B total / 3B active vs 27.78B
Qwen3.6-35B vs Qwen3.6 27B
Qwen3.6-35B is 35B total / 3B active, 262,144 tokens on the official card. Qwen3.6 27B is 27.78B, 262,144 tokens. Q6 / Q5 / Q4 file sizes come from the listed GGUFs on each /models page, or a calculator estimate when a listing is missing. Q3 stays off. Q4 files fit one R9700 at 4K on the calculator for both sides. Qwen3.6-35B has measured llama.cpp cells on the local desktop. Qwen3.6 27B speed is not measured yet. llama.cpp first.
-
125B total / 6B active vs 320B total / 18B active
Qwen 3.8 Flash vs GLM 5.3 Flash
Qwen 3.8 Flash is 125B total / 6B active, 262,144 tokens on the official card. GLM 5.3 Flash is 320B total / 18B active, 1,048,576 tokens. Q6 / Q5 / Q4 file sizes come from the listed GGUFs on each /models page, or a calculator estimate when a listing is missing. Q3 stays off. Q4 does not sit in one R9700 VRAM on either side — expect RAM offload. Neither model was measured on the local desktop. Fit numbers are calculator estimates. llama.cpp first.
-
27.4B vs 27.78B
Gemma 3 27B vs Qwen3.6 27B
Gemma 3 27B is 27.4B, 128K tokens on the official card. Qwen3.6 27B is 27.78B, 262,144 tokens. Q6 / Q5 / Q4 file sizes come from the listed GGUFs on each /models page, or a calculator estimate when a listing is missing. Q3 stays off. Q4 files fit one R9700 at 4K on the calculator for both sides. Neither model was measured on the local desktop. Fit numbers are calculator estimates. llama.cpp first.
-
27.4B vs 27.8B
Gemma 3 27B vs Qwen3.8 27B
Gemma 3 27B is 27.4B, 128K tokens on the official card. Qwen3.8 27B is 27.8B, 262,144 tokens. Q6 / Q5 / Q4 file sizes come from the listed GGUFs on each /models page, or a calculator estimate when a listing is missing. Q3 stays off. Q4 files fit one R9700 at 4K on the calculator for both sides. Neither model was measured on the local desktop. Fit numbers are calculator estimates. llama.cpp first.
-
27.78B vs 27.8B
Qwen3.6 27B vs Qwen3.8 27B
Qwen3.6 27B is 27.78B, 262,144 tokens on the official card. Qwen3.8 27B is 27.8B, 262,144 tokens. Q6 / Q5 / Q4 file sizes come from the listed GGUFs on each /models page, or a calculator estimate when a listing is missing. Q3 stays off. Q4 files fit one R9700 at 4K on the calculator for both sides. Neither model was measured on the local desktop. Fit numbers are calculator estimates. llama.cpp first.
-
70.6B vs 320B total / 18B active
Llama 3.3 70B vs GLM 5.3 Flash
Llama 3.3 70B is 70.6B, 128K tokens on the official card. GLM 5.3 Flash is 320B total / 18B active, 1,048,576 tokens. Q6 / Q5 / Q4 file sizes come from the listed GGUFs on each /models page, or a calculator estimate when a listing is missing. Q3 stays off. Q4 does not sit in one R9700 VRAM on either side — expect RAM offload. Neither model was measured on the local desktop. Fit numbers are calculator estimates. llama.cpp first.
-
21.5B total / 3.6B active vs 27.4B
gpt-oss 20B vs Gemma 3 27B
gpt-oss 20B is 21.5B total / 3.6B active, 131,072 tokens on the official card. Gemma 3 27B is 27.4B, 128K tokens. Q6 / Q5 / Q4 file sizes come from the listed GGUFs on each /models page, or a calculator estimate when a listing is missing. Q3 stays off. Q4 files fit one R9700 at 4K on the calculator for both sides. Neither model was measured on the local desktop. Fit numbers are calculator estimates. llama.cpp first.
Every cited model vs every other
Works with JavaScript disabled. Each heading is one model page; the links are static pair pages.
Qwen3.6-35B 35B total / 3B active · 262,144 tokens
Qwen 3.8 Flash 125B total / 6B active · 262,144 tokens
gpt-oss 20B 21.5B total / 3.6B active · 131,072 tokens
Gemma 3 27B 27.4B · 128K tokens
Llama 3.3 70B 70.6B · 128K tokens
GLM 5.3 Flash 320B total / 18B active · 1,048,576 tokens
Qwen3.6 27B 27.78B · 262,144 tokens
Qwen3.8 27B 27.8B · 262,144 tokens
Stack Lab
Doctor is allowlisted on every compare. House daily / Flash / agent profiles stay on those house model pages only.
If the app is not installed, packages are not published yet. There is no public download URL. Read the Stack Lab install notes . If Stack Lab is already running on this machine, open the local app .
Also: Compare 35B · Compare GPUs · Models · Benchmarks · Calculator · Stack Lab · Qwen3.6-35B · Qwen 3.8 Flash · GLM 5.3 Flash