Sunk Cost sunkcost.ai Data checked 2026-09-03

Every open model, measured against the frontier

48 open-weight models you can download and run at home, ranked on the Artificial Analysis Intelligence Index v4.3, with the hosted models from Anthropic and OpenAI dropped into the same table for scale. Each row links to what it takes to run it.

The short version: the best open model here scores 42 — that is GLM-5.3-Flash, and it wants 189 GB of memory. The best hosted model scores 53. That gap of 11 points is the thing no amount of hardware closes.

ModelScoreClassGood atWeightsCheapest machine that runs itNext down
GPT-6 Astrahosted 53 Runs in someone else’s data centre. You cannot download it. —
Claude Fable 5.1hosted 53 Runs in someone else’s data centre. You cannot download it. —
Claude Opus 5hosted 51 Runs in someone else’s data centre. You cannot download it. —
GPT-5.6 Solhosted 47 Runs in someone else’s data centre. You cannot download it. —
GPT-5.6 Terrahosted 42 Runs in someone else’s data centre. You cannot download it. —
Claude Sonnet 5hosted 38 Runs in someone else’s data centre. You cannot download it. —
GPT-5.6 Lunahosted 38 Runs in someone else’s data centre. You cannot download it. —
Claude Haiku 4.5hosted 18 Runs in someone else’s data centre. You cannot download it. —
GLM-5.3-FlashUD-Q4_K_M 42 Sonnet-class 189 GB Mac Studio M5 Ultra, 256GB $10,799 vs Qwen3.8 Flash Next
Qwen3.8 Flash NextQ4_K_M 40 Sonnet-class 120 GB Mac Studio M5 Ultra, 256GB $10,799 vs DeepSeek V4-Flash
DeepSeek V4-FlashUD-Q4_K_M 35 Sonnet-class 155 GB Mac Studio M5 Ultra, 256GB $10,799 vs Qwen3.8 27B
Qwen3.8 27BQ4_K_M 34 Sonnet-class 16 GB Strix Halo Framework Desktop, 32GB $1,269 vs Inkling Small
Inkling SmallUD-Q4_K_M 26 Haiku-class 163 GB Mac Studio M5 Ultra, 256GB $10,799 vs Tencent Hy3
Tencent Hy3Q4_K_M 26 Haiku-class 182 GB nothing on the list vs Ling 3.0 flash
Ling 3.0 flashQ4_K_M 25* Haiku-class 78 GB Strix Halo Framework Desktop, 128GB $3,449 vs MiniMax M2.7
MiniMax M2.7Q4_K_M 23 Haiku-class 140 GB Mac Studio M5 Ultra, 256GB $10,799 vs Qwen3.6 27B
Qwen3.6 27BQ4_K_M 22 Haiku-class 17 GB Strix Halo Framework Desktop, 32GB $1,269 vs Qwen3.6 35B-A3B
Qwen3.6 35B-A3BQ4_K_M 19 Haiku-class 22 GB Strix Halo Framework Desktop, 32GB $1,269 vs Muse Glimmer 30B
Muse Glimmer 30BQ4_K_M 18 Haiku-class 17 GB Strix Halo Framework Desktop, 32GB $1,269 vs Gemma 4 26B-A4B
Gemma 4 26B-A4BQ4_K_M 17* Haiku-class 17 GB Strix Halo Framework Desktop, 32GB $1,269 vs Qwen3.5 122B-A10B
Qwen3.5 122B-A10BUD-Q4_K_M 16 Haiku-class 78 GB Strix Halo Framework Desktop, 128GB $3,449 vs Gemma 4 31B it
Gemma 4 31B itQ4_K_M 15 Haiku-class 20 GB Strix Halo Framework Desktop, 64GB $1,959 vs Granite 4.2 30B
Granite 4.2 30BQ4_K_M 15* Haiku-class 18 GB Strix Halo Framework Desktop, 64GB $1,959 vs GLM-4.7-Flash
GLM-4.7-FlashQ4_K_M 15* Haiku-class 18 GB Strix Halo Framework Desktop, 32GB $1,269 vs Gemma 4 12B
Gemma 4 12BQ4_K_M 14* Below every hosted tier 7.1 GB Mac mini M6, 16GB $899 vs Qwen3.5 9B
Qwen3.5 9BQ4_K_M 14* Below every hosted tier 5.7 GB Mac mini M6, 16GB $899 vs Nemotron 3.5 Lightning 30B-A3B
Nemotron 3.5 Lightning 30B-A3BQ4_K_M 14 Below every hosted tier 25 GB Strix Halo Framework Desktop, 64GB $1,959 vs Qwen3 235B-A22B Instruct 2507
Qwen3 235B-A22B Instruct 2507Q4_K_M 13 Below every hosted tier 142 GB Mac Studio M5 Ultra, 256GB $10,799 vs MiniCPM5 2B
MiniCPM5 2BQ4_K_M 13 Below every hosted tier 1.6 GB Mac mini M6, 16GB $899 vs Qwen3.5 4B
Qwen3.5 4BQ4_K_M 13* Below every hosted tier 2.7 GB Mac mini M6, 16GB $899 vs gpt-oss-120b
gpt-oss-120bMXFP4 12 Below every hosted tier 63 GB Strix Halo Framework Desktop, 128GB $3,449 vs Ling 3.0 tiny
Ling 3.0 tinyQ4_K_M 12 Below every hosted tier 4.8 GB Mac mini M6, 16GB $899 vs Granite 4.2 8B
Granite 4.2 8BQ4_K_M 12 Below every hosted tier 5.3 GB Mac mini M6, 24GB $1,099 vs GLM-4.5-Air
GLM-4.5-AirQ4_K_M 11* Below every hosted tier 73 GB Strix Halo Framework Desktop, 128GB $3,449 vs Mistral Small 4 (119B-2603)
Mistral Small 4 (119B-2603)Q4_K_M 11 Below every hosted tier 74 GB Strix Halo Framework Desktop, 128GB $3,449 vs Qwen3 30B-A3B Instruct 2507
Qwen3 30B-A3B Instruct 2507Q4_K_M 10 Below every hosted tier 19 GB Strix Halo Framework Desktop, 32GB $1,269 vs Qwen3-Coder 30B-A3B
Qwen3-Coder 30B-A3BQ4_K_M 10* Below every hosted tier 19 GB Strix Halo Framework Desktop, 32GB $1,269 vs Qwen3-Coder Next
Qwen3-Coder NextQ4_K_M 10 Below every hosted tier 48 GB Strix Halo Framework Desktop, 128GB $3,449 vs gpt-oss-20b
gpt-oss-20bMXFP4 9 Below every hosted tier 12 GB Mac mini M6, 24GB $1,099 vs Gemma 4 E4B
Gemma 4 E4BQAT Q4_0 9* Below every hosted tier 5.2 GB Mac mini M6, 16GB $899 vs Devstral 2 123B
Devstral 2 123BQ4_K_M 9 Below every hosted tier 75 GB Strix Halo Framework Desktop, 128GB $3,449 vs DeepSeek-R1-Distill-Qwen-32B
DeepSeek-R1-Distill-Qwen-32BQ4_K_M 8* Below every hosted tier 20 GB Strix Halo Framework Desktop, 64GB $1,959 vs DeepSeek-R1-Distill-Llama-70B
DeepSeek-R1-Distill-Llama-70BQ4_K_M 8* Below every hosted tier 43 GB Strix Halo Framework Desktop, 128GB $3,449 vs Llama 3.3 70B Instruct
Llama 3.3 70B InstructQ4_K_M 8* Below every hosted tier 43 GB Strix Halo Framework Desktop, 128GB $3,449 vs LFM2.5 2.6B
LFM2.5 2.6BQ4_K_M 8* Below every hosted tier 1.7 GB Mac mini M6, 16GB $899 vs Devstral Small 2 24B
Devstral Small 2 24BQ4_K_M 8 Below every hosted tier 14 GB Strix Halo Framework Desktop, 32GB $1,269 vs Llama 3.1 8B Instruct
Llama 3.1 8B InstructQ4_K_M 7* Below every hosted tier 4.9 GB Mac mini M6, 16GB $899 vs Qwen3 32B
Qwen3 32BQ4_K_M 7 Below every hosted tier 20 GB Strix Halo Framework Desktop, 64GB $1,959 vs Mistral Small 3.2 24B Instruct
Mistral Small 3.2 24B InstructQ4_K_M 7 Below every hosted tier 14 GB Strix Halo Framework Desktop, 32GB $1,269 vs Qwen3 14B
Qwen3 14BQ4_K_M 6 Below every hosted tier 9.0 GB Mac mini M6, 24GB $1,099 vs Llama 4 Scout 17B-16E
Llama 4 Scout 17B-16EQ4_K_M 6 Below every hosted tier 65 GB Strix Halo Framework Desktop, 128GB $3,449 vs Ministral 3 14B
Ministral 3 14BQ4_K_M 6 Below every hosted tier 8.2 GB Mac mini M6, 24GB $1,099 vs Qwen3 8B
Qwen3 8BQ4_K_M 5 Below every hosted tier 5.0 GB Mac mini M6, 16GB $899 vs Gemma 3 27B it
Gemma 3 27B itQ4_K_M 5 Below every hosted tier 17 GB Strix Halo Framework Desktop, 32GB $1,269 vs Ministral 3 8B
Ministral 3 8BQ4_K_M 5 Below every hosted tier 5.2 GB Mac mini M6, 16GB $899 vs Gemma 3 12B it
Gemma 3 12B itQ4_K_M 4 Below every hosted tier 7.3 GB Mac mini M6, 16GB $899

Marked 'estimated' where Artificial Analysis imputed the score rather than running the full v4.3 suite. 26 of the models here carry that mark. Scores are the Artificial Analysis Intelligence Index v4.3 (source), read on 2026-09-11. Hybrid models are shown at their reasoning or highest-effort score, with the alternative noted on each model's page. The dots are, in order: summarising, translation, everyday coding, reasoning & maths, agentic work.