Models
The same prompt does not land the same way on every model, because the vendors document different preferences. These pages summarise what each one asks for, and link to the documentation it came from.
OpenAI
Rewards restraint. The current generation performs better with shorter, less repetitive prompts than its predecessors did.
Anthropic
The one family with a documented preference for XML structure, and the one that responds most to being told why an instruction matters.
Prefers short over persuasive. Gemini 3 answers direct instructions better than elaborate ones, and puts your question last when there is a lot of data.
Meta
Open weights, and historically the family that made self-hosting mainstream. Worth understanding — but Llama 4 has not been updated since May 2025, so treat it as a known quantity rather than a current recommendation.
Microsoft
The only one of the five that can already see your files. That changes what a good prompt looks like: less structure, more pointing.
xAI
Very large context and a high reasoning ceiling — but almost no published prompting guidance for its text models, so treat advice about Grok with more caution than the rest.
Mistral AI
The best prompting documentation of any provider here, an EU inference region, and almost the whole line-up available as open weights.
DeepSeek
The cheapest option here by a wide margin and the only one that open-weights its actual frontier model — with the caveat that the hosted API stores data in China.
Compare all models
| Provider | Models | Context | Max output | Good for |
|---|---|---|---|---|
| OpenAI | GPT-5.6 Sol Flagship for complex professional work. | 1.05M | 128K | Hard analysis, long documents, agentic workflows with many tool calls. |
| OpenAI | GPT-5.6 Terra Balances intelligence and cost. | 1.05M | 128K | The sensible default for most drafting, summarising and reasoning work. |
| OpenAI | GPT-5.6 Luna Built for budget-conscious, high-volume work. | 1.05M | 128K | Classification, extraction, routing, and anything you run thousands of times. |
| Anthropic | Claude Fable 5 Next-generation intelligence for long-running agents. | 1M | 128K | The hardest reasoning and long-horizon agent work. |
| Anthropic | Claude Opus 5 For complex agentic coding and enterprise work. | 1M | 128K | Coding, analysis, and multi-step professional work at moderate latency. |
| Anthropic | Claude Sonnet 5 The best combination of speed and intelligence. | 1M | 128K | Day-to-day drafting, summarising and interactive use. |
| Anthropic | Claude Haiku 4.5 The fastest model with near-frontier intelligence. | 200K | 64K | High-volume classification, extraction and quick turns. |
| Gemini 3.7 Flash Built for complex coding, agentic workflows and reliable multi-step execution. | — | — | Agentic and coding work where you want speed alongside multi-step reliability. | |
| Gemini 3.6 Flash Balances speed and multimodal capability for general agentic tasks. | — | — | Mixed text-and-image work at interactive speed. | |
| Gemini 3.5 Flash-Lite The fastest, most cost-effective model in its generation for high-throughput work. | — | — | Classification, tagging, and very high request volumes. | |
| Gemini 2.5 Pro Previous-generation flagship for complex reasoning and coding. | — | — | Workloads already tuned against it, and cases where you want the older, more verbose default. | |
| Meta | Llama 4 Scout 17B active parameters across 16 experts (109B total), with a 10M-token context window. | 10M | — | Very long inputs — full codebases, archives, transcripts — on hardware you control. |
| Meta | Llama 4 Maverick 17B active parameters across 128 experts (400B total). | — | — | General assistant and reasoning work where you want the stronger open model. |
| Microsoft | Microsoft 365 Copilot Assistant embedded across Word, Excel, Outlook, Teams and PowerPoint, grounded in your tenant’s content. | — | — | Summarising meetings and threads, drafting from your own documents, and finding things across the tenant. |
| xAI | Grok 4.6 Frontier model for coding, agentic tasks and knowledge work. | 500K | [object Object] | Hard reasoning and agentic work where you also want very long output. |
| xAI | Grok 4.3 Efficient long-context alternative. | 1M | — | Very long documents at lower cost than the flagship. |
| xAI | Grok 4.20 Available as reasoning, non-reasoning and multi-agent variants. | 1M | — | Choosing explicitly whether you want reasoning at all, which the frontier models do not allow. |
| Mistral AI | Mistral Large 3 State-of-the-art, open-weight, general-purpose multimodal model. | 256K | — | General work where you want frontier-class output at a low price, or the option to self-host the same model. |
| Mistral AI | Mistral Medium 3.5 Frontier-class multimodal model for agentic and coding use. | 256K | — | Agentic and coding workloads, with documented strong adherence to system prompts. |
| Mistral AI | Mistral Small 4 Hybrid model unifying instruct, reasoning and coding in one. | 256K | — | The value pick, and the one most people should start with. |
| Mistral AI | Ministral 3 8B Efficient small model with text and vision. | 256K | — | High volume, low cost, and running locally — it is about 5 GB quantised. |
| DeepSeek | DeepSeek V4 Pro 1.6T total parameters with 49B active, the more capable tier. | 1M | 384K | Hard work at a fraction of competitors’ prices, and very long generations. |
| DeepSeek | DeepSeek V4 Flash 284B total parameters with 13B active, the fast tier. | 1M | 384K | High volume at the lowest price of anything on this site, with the same context and output limits as the larger model. |
A dash means the vendor does not publish that figure in a comparable place. Open the provider page and follow its documentation link for current specifications.