Reference — Models
The right model for the job.
The gap between open-weight and frontier models has nearly closed. The question is no longer "can open-weight compete?" It's "which model fits your workload, your compliance requirements, and your budget."
Quick pick
Start here.
IBM Granite 4.1
Apache 2.0, on-premises, US-made. ISO 42001 certified and cryptographically signed — the compliance story regulators want.
Claude Fable 5
Via API. Adaptive reasoning, 1M-token context. Pay per use.
Nemotron 3 · Llama 4
Hosted dedicated. US open-weight families you own outright — strong multi-step reasoning at scale. (GLM-5.2 leads among open-weight models if origin isn't a constraint.)
Closed models
Frontier models (API-only).
Most intelligent. Least control. Pay per token.
| Model | Provider | Best for | Context |
|---|---|---|---|
| Claude Fable 5 | Anthropic | Deepest reasoning, complex coding, adaptive thinking | 1M |
| GPT-5.6 Sol | OpenAI | General purpose, math, broad ecosystem | 1M (est.) |
| Gemini 3.1 Pro | Multimodal, video, long context | 1M | |
| Grok 4.5 | xAI | Real-time data, strong general reasoning | 500K |
Pricing changes monthly — we quote current rates in every proposal
Trade-off:Maximum intelligence. Your data transits the vendor's infrastructure on every API call. The enterprise agreements we configure prevent training on your data.
Open models
Open-weight models (self-hostable).
Full control. Self-host or deploy on dedicated infrastructure.
| Model | Lab | Parameters | License | Best for |
|---|---|---|---|---|
| Llama 4 | Meta · US | MoE family | Llama 4 Community | Long context (up to 10M), broad ecosystem |
| Gemma 4 | Google · US | 2B–31B dense + 26B MoE | Apache 2.0 | Multimodal, 140+ languages, efficient edge-to-server |
| Nemotron 3 | NVIDIA · US | 30B–550B (MoE) | NVIDIA Open Model | Plans and runs multi-step tasks; tuned for NVIDIA GPUs |
| Granite 4.1 | IBM · US | 3B–30B dense | Apache 2.0 | Enterprise-grade; ISO 42001 certified; tamper-proof (cryptographically signed) |
| Phi-4 | Microsoft · US | ~14B (+ mini) | MIT | Efficient reasoning; runs on a laptop or CPU |
| OLMo 3 | Allen Institute · US | 7B–32B | Apache 2.0 | Fully open (weights + data + code) — maximum auditability |
| Mistral Large 3 | Mistral · France | 675B (MoE) | Apache 2.0 | Unrestricted license, multilingual |
| DeepSeek V4 Pro | DeepSeek · China | 1.6T (MoE) | MIT | Reasoning & coding leader; lowest cost floor |
| GLM-5.2 | Zhipu · China | 754B (MoE) | MIT | Highest-scoring open-weight model on public tests |
| Kimi K2.6 | Moonshot · China | ~1T (MoE) | Modified MIT | Agentic coding, long-horizon tool use (256K context) |
| Qwen 3.5 | Alibaba · China | 235B–397B (MoE) | Apache 2.0 | Broad benchmark strength, multilingual |
MoE (Mixture-of-Experts): the model activates only the parts it needs for each request — large capacity, lower running cost.
Trade-off: the best open-weight models now sit within a few percent of frontier on public leaderboards (~8% behind in early 2024, ~2–3% today) and lag roughly four months — though closed models keep an edge at the very top of reasoning and agentic-coding tasks. Zero data leakage. Full audit trail. You own the model. For most business tasks, open-weight is production-ready.
Origin matters for some buyers. For organizations that prefer to avoid models of foreign origin, there is now a strong US-made open-weight slate — Llama, Gemma, Nemotron, Granite, and Phi, plus the fully-open OLMo — alongside the leading Chinese labs. We match the model to your policy, not the other way around.
Legal
Licenses explained.
| License | Commercial use? | Can modify? | Must share changes? |
|---|---|---|---|
| MIT (DeepSeek, GLM, Phi) | Yes | Yes | No |
| Apache 2.0 (Gemma, Granite, OLMo, Qwen, Mistral) | Yes | Yes | No |
| Modified MIT (Kimi) | Yes (UI attribution above 100M MAU / $20M rev) | Yes | No |
| Llama 4 Community / NVIDIA Open | Yes (with limits) | Yes | No |
| Proprietary | Via API only | No | N/A |
For regulated enterprises that need legal certainty, we recommend Apache 2.0 models — no usage caps, no license surprises.
Framework
Intelligence vs. Control vs. Cost.
More control = open-weight on your hardware. More intelligence = frontier via API. The best setup is often both: open-weight for volume, frontier for hard problems.
Model landscape last verified July 8, 2026. This changes fast — new models are released weekly. Talk to us for current recommendations.
Contact
Not sure which model fits?
We'll map your use case to the right model — and the right deployment mode.
Get a model recommendation →