Model finder
Find a model
Describe the need; the atlas applies deterministic rules to observed attributes, current prices and current benchmark results and returns the models that satisfy them — sorted by how many criteria they meet, then best benchmark rank, then release date. No composite score, no single winner.
11 models satisfy at least one of the criteriaNo single winner — these are the observed dimensions; the order is criteria met → best rank → release date.
use case codingdeployment localmemory gb 64quant 4bit
| Model | Why | Observed dimensions | Fit Estimated | Ranks | |
|---|---|---|---|---|---|
| Qwen3.8 27BOpen weightsQwen |
| 27.8B params · 1M context · Apache-2.0 · image · text · video · reasoning · tool calling · released 5 Aug 2026 · from $0.214 in / $2.55 out · 2 providers | ✓ fits16.5 GB+45.5 GB headroom4bit | livebench-agentic-coding#11livebench-if#16livebench-data-analysis#24+10 | |
| Qwen3.6 35B A3BOpen weightsQwen |
| 36B params · 262.1K context · Apache-2.0 · image · text · video · reasoning · tool calling · released 15 Apr 2026 · from $0.10 in / $0.90 out · 1 provider | ✓ fits21.2 GB+40.8 GB headroom4bit | tau2-bench#22mmmu-pro#45terminal-bench#59+5 | |
| Gemma 4 31BOpen weightsGoogle |
| 31.3B params · 262.1K context · Apache-2.0 · image · text · video · reasoning · tool calling · released 11 Mar 2026 · from $0.09 in / $0.34 out · 3 providers | ✓ fits18.5 GB+43.5 GB headroom4bit | ifbench#28terminal-bench#51scicode#55+5 | |
| Qwen3.6 27BOpen weightsQwen |
| 27.8B params · 262.1K context · Apache-2.0 · image · text · video · reasoning · tool calling · released 21 Apr 2026 · from $0.30 in / $2 out · 2 providers | ✓ fits16.5 GB+45.5 GB headroom4bit | tau2-bench#31livebench-data-analysis#42livebench-coding#46+13 | |
| Qwen3 32BOpen weightsQwen |
| 32.8B params · 131.1K context · Apache-2.0 · text · reasoning · tool calling · released 27 Apr 2025 · from $0.08 in / $0.28 out · 1 provider | ✓ fits19.3 GB+42.7 GB headroom4bit | aider-polyglot#37aider-polyglot-well-formed#57scicode#100+6 | |
| QwQ-32BOpen weightsAlibaba Group |
| 32B params · 131.1K context · Apache-2.0 · text · reasoning · tool calling · released 6 Mar 2025 · 0 providers | ✓ fits18.9 GB+43.1 GB headroom4bit | aider-polyglot#49aider-polyglot-well-formed#60ifbench#211+3 | |
| Nemotron 3 Nano 30B A3BOpen weightsNVIDIA |
| 31.6B params · 262.1K context · Other · text · reasoning · tool calling · released 4 Dec 2025 · from $0.05 in / $0.20 out · 2 providers | ✓ fits18.7 GB+43.3 GB headroom4bit | ifbench#57scicode#113terminal-bench#150+4 | |
| North Mini Code (free)Open weightsCohere |
| 30.5B params · 256K context · Apache-2.0 · reasoning · released 5 Jun 2026 · 2 providers | ✓ fits18.0 GB+44.0 GB headroom4bit | terminal-bench#79scicode#87ifbench#114+4 | |
| gpt-oss-20bOpen weightsOpenAI |
| 20.9B params · 131.1K context · Apache-2.0 · text · reasoning · tool calling · released 4 Aug 2025 · from $0.03 in / $0.13 out · 3 providers | ✓ fits12.5 GB+49.5 GB headroom4bit | scicode#86ifbench#88tau2-bench#131+4 | |
| Qwen3 8BOpen weightsQwen |
| 8.19B params · 131.1K context · Apache-2.0 · text · reasoning · tool calling · released 27 Apr 2025 · from $0.117 in / $0.455 out · 1 provider | ✓ fits5.2 GB+56.8 GB headroom4bit | scicode#115tau2-bench#202terminal-bench#240+4 | |
| LFM2.5-2.6B (free)Open weightsLiquid AI |
| 2.7B params · 128K context · Other · reasoning · released 28 Jul 2026 · 1 provider | ✓ fits2.0 GB+60.0 GB headroom4bit | scicode#129humanitys-last-exam#266artificial-analysis-intelligence-index#279+1 |
Green chips = criteria satisfied by an observed fact; dashed amber = satisfied by an estimate (hardware fit). Ranks are positions inside each benchmark's primary comparability group.
MethodDeterministic filters over observed attributes, current prices and current benchmark results of canonical models; sorted by number of satisfied criteria, then best benchmark rank, then release date. No composite score. Hardware fit is an estimate. /methodology