The most important open-weight models in June 2026. Which model fits which , what hardware requirement is behind it, which licence is in play. Updated with every new release.
Pull models via Ollama
# All-purpose hero, runs on Apple Silicon M4 Ultra (192 GB) or NVIDIA with 48+ GB VRAM
ollama pull llama3.3:70b
# Excellent multilingual (esp. German), fits an RTX 5090 (32 GB)
ollama pull qwen2.5:32b
# Mixture-of-Experts, very strong tokens-per-dollar
ollama pull mixtral:8x7b
# Reasoning specialist (distill variant for realistic hardware)
ollama pull deepseek-r1:32b
# Small and strong, feasible on a single RTX 4090 (24 GB)
ollama pull phi4:14b
ollama pull gemma3:27b
# OpenAI's open-weight model (released Aug 2025), 20B Q4 needs ~12 GB VRAM
ollama pull gpt-oss:20bEvery model has its own strengths — size, licence, context length, . Compact overview here, clickable to the official vendor model page.
Meta
Meta flagship
Meta's all-purpose hero. Very balanced at text generation, knowledge tasks, multilingual work. Llama 3.3 Community License allows commercial SMB use.
ollama pull llama3.3:70b
Mistral AI
French powerhouse
Mistral 7B as a compact model, Mixtral 8x7B/8x22B as a Mixture-of-Experts with a very strong performance/VRAM ratio. All Apache-2.0 — fully commercially usable.
ollama pull mixtral:8x7b
Alibaba
Multilingual specialist
Alibaba model with excellent German quality, even at smaller sizes. Variants from 0.5B to 72B. Apache-2.0 — unproblematic even for GDPR-sensitive SMBs.
ollama pull qwen2.5:32b
Google small model
3 in sizes from 1B to 27B — very efficient for its weight class. License (similar to Llama: commercial use possible, some obligations).
ollama pull gemma3:27b
DeepSeek
Reasoning specialist
DeepSeek R1 sets the bar for logical reasoning and mathematics. Full 671B version only on enterprise hardware, distill variants 7B/14B/32B/70B usable on SMB hardware. MIT license.
ollama pull deepseek-r1:32b
Microsoft
Microsoft small champion
Phi 4 with only 14B parameters shows: well-curated training data beats sheer size. Very good performance on a single RTX 4090. MIT license, commercially unrestricted.
ollama pull phi4:14b
OpenAI
OpenAI's open-weight model
In August 2025 OpenAI released gpt-oss in two sizes (20B and 120B) under Apache-2.0. In model quality clearly better than older OSS models, with full commercial release.
ollama pull gpt-oss:20b
An open-weight model is an whose trained weights (the parameter file) are publicly available — typically as a download on HuggingFace, on the vendor website or through . That distinguishes them from ChatGPT, Claude or Gemini Pro, whose weights are not released and which are accessible only via APIs.
Important: 'open weight' is not the same as 'open source'. The licence variation is wide — from real MIT/Apache-2.0 (Mistral, Qwen, Phi, GPT-oss) through custom licences with branding clauses (Llama, ) to restricted 'research only' licences. For commercial SMB use every licence is to be reviewed in advance — footnote per model on the respective card.
Recommendations from our own client practice. Not 'the best model' — there is no best. But 'for this this model is the most pragmatic'.
Six properties that make these seven models a class of their own — and distinguish them from cloud frontier models.
OpenAI-compatible API call against a local model
# Pick a model (variable)
MODEL="llama3.3:70b"
# Alternative: qwen2.5:32b · mixtral:8x7b · deepseek-r1:32b · gpt-oss:20b
# OpenAI-compatible chat-completion call
curl http://localhost:11434/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "'$MODEL'",
"messages": [
{"role": "system", "content": "You are a pragmatic AI consultant for SMBs."},
{"role": "user", "content": "Which model do you recommend for an 8-person law firm with an RTX 4090?"}
],
"temperature": 0.2
}'
# The same request also works with the OpenAI Python SDK,
# simply set OPENAI_API_BASE to http://localhost:11434/v1.Cloud alternatives honestly compared
Three cloud frontier models that are still often ahead in absolute quality. Trade-off: data flow and per-token costs against top quality.
Frontier cloud (USA)
OpenAI
Frontier cloud (USA/EU)
Anthropic
Frontier cloud (USA)
Rule of thumb: with GDPR obligations or IP protection — open-weight local. For maximum quality on uncritical content — cloud frontier. For both — multi-routing in (sensitive content local, general research cloud).
Pricing
License
Mix of Apache-2.0 (Mistral, Qwen, Phi, GPT-oss), MIT (DeepSeek), custom community licences (Llama, Gemma). For SMB own-use in nearly all constellations unproblematic — verify per model on the respective card.
Hardware costs
Mid-range (for models up to 32B): RTX 4090 24GB + 32GB RAM workstation, around €3,000. Premium (for 70B+): Apple Mac Studio M4 Ultra 192 GB from €8,000 or an NVIDIA server with RTX 5090. Power: 150–300 W during inference, <50 W idle.
vs. cloud API
Cloud costs scale linearly with volume. At 1,000 tokens/day per 10 staff: GPT-5 around €60/month, Claude Sonnet similar. Self-hosted: hardware amortisation in 12–18 months. From 50+ staff clearly cheaper.
Important: quantisation quality is not linear. Q4_K_M is standard and 'good enough' for 95 % of . Q5/Q6 for higher demands. Full FP16 only for research. Anyone trying to make do with less VRAM should first check quantisation before picking a smaller model.
Related topics
Models alone are useless. loads them, provides the user surface, the server solution sits behind:
Free intro call, no strings attached. In 30 minutes you'll know whether and how AI can help your business.