FITMYLLM · JULY 28, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
Quantization Options Select your GPU for speed estimates Context length: 4K 8K 16K 32K 64K 128K 256K 512K 1M
Quant Bits VRAM @ 16K Quality Q3_K_M 4 12.6 GB
10.4 + 2.3 KV
low Q3_K_L 4.3 13.4 GB
11.1 + 2.3 KV
moderate IQ4_XS 4.46 13.8 GB
11.5 + 2.3 KV
moderate Q4_K_S 4.67 14.3 GB
12.0 + 2.3 KV
moderate Q4_K_M 4.89 14.8 GB
12.6 + 2.3 KV
good Q5_K_S 5.57 16.5 GB
14.3 + 2.3 KV
good Q5_K_M 5.7 16.8 GB
14.6 + 2.3 KV
good Q6_K 6.56 19.0 GB
16.7 + 2.3 KV
excellent Q8_0 8.5 23.8 GB
21.5 + 2.3 KV
lossless FP16 16 42.3 GB
40.1 + 2.3 KV
lossless
Select your GPU above to see speed estimates and compatibility for each quantization.
Deploying for a team or in production? Size GPUs, cost & scaling in Enterprise → ▸ READY TO RUN THIS? RENT BY THE HOUR
RENT A GPU AND RUN INTERNLM2.5 20B NOW
Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.
Community Ratings Chat Coding Reasoning Creative Vision Roleplay Agentic
Loading ratings...
Run this model Q3_K_M — 10.4 GB VRAM Q3_K_L — 11.1 GB VRAM IQ4_XS — 11.5 GB VRAM Q4_K_S — 12.0 GB VRAM Q4_K_M — 12.6 GB VRAM Q5_K_S — 14.3 GB VRAM Q5_K_M — 14.6 GB VRAM Q6_K — 16.7 GB VRAM Q8_0 — 21.5 GB VRAM FP16 — 40.1 GB VRAM
Ollama llama.cpp vLLM LM Studio KoboldCpp Jan Docker
▸ Easiest way to get started · Beginners
DOCS ↗ curl -fsSL https://ollama.com/install.sh | shCOPY
$ ollama run internlm2:20b-chat-v2.5-q4_K_MCOPY
Downloads and runs automatically. Add --verbose for speed stats.
▸ SETUP GUIDE >_
Auto-setup with fitmyllm CLI Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.
Auto-detect GPU Live tok/s in chat Speed benchmarks 9 inference engines
GPUs that can run this model At Q4_K_M quantization. Sorted by minimum VRAM.
Find the best GPU for InternLM2.5 20B
Build Hardware for InternLM2.5 20B InternLM2.5 20B — latest version with improved reasoning and math.
Read full model card ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
InternLM2.5 20B — 19.8B Dense. ▸ SPECIFICATIONS
PARAMETERS 19.8B
ARCHITECTURE Dense Transformer
CONTEXT LENGTH 1024K tokens
CAPABILITIES chat, coding, reasoning
RELEASE DATE 2024-08-05
PROVIDER Shanghai AI Lab
FAMILY internlm ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY Q3_K_M 4 10.4 GB 88% Q3_K_L 4.3 11.1 GB 90% IQ4_XS 4.46 11.5 GB 92% Q4_K_S 4.67 12.0 GB 93% Q4_K_M 4.89 12.6 GB 94% Q5_K_S 5.57 14.3 GB 96% Q5_K_M 5.7 14.6 GB 96% Q6_K 6.56 16.7 GB 97% Q8_0 8.5 21.5 GB 100% FP16 16 40.1 GB 100%
§ 01 BENCHMARK SCORES
HumanEval 72.0
MMLU-PRO 52.0
MATH 68.0
IFEval 75.0
BBH 62.8
GPQA 9.5
MUSR 16.7
Arena Elo 1164.0
§ 02 RUN COMMAND
Run InternLM2.5 20B locally with Ollama — needs 12.6 GB VRAM at Q4_K_M:
$ ollama run internlm2:20b
§ 03 COMPATIBLE GPUs
30 @ Q4_K_M Feedback 