Explore the models
Filter by tier, search by name or job, tap any card to open the full write-up.
The deepest reasoning I have access to. Slow, expensive, and worth it when the output has to survive scrutiny.
- Governance and risk modelling
- Proposals that face objections
- Complex multi-step architecture
Overkill for everyday drafting, and you will feel the latency.
Broad, reliable heavyweight reasoning with a calmer tone than most flagships.
- Long document analysis
- Hard refactors
- Strategy work
Cost adds up fast if you route routine work here.
The best taste of any model I use. Give it an interface and it comes back looking designed.
- Front-end and UI generation
- Landing page copy with voice
- Creative direction
Confident on business logic even when it is wrong. Review the reasoning, not just the look.
The strongest multimodal model: video, audio, and huge context in one place.
- Watching recorded meetings
- Massive context dumps
- Spreadsheet and data review
Writing voice is flatter than Claude or GPT. Fine for analysis, weak for final copy.
Fast flagship with live access to public social conversation.
- Real-time trend checks
- Blunt second opinions
Tone needs editing before anything client-facing.
The best balance on the market right now: near-flagship quality at working speed.
- Daily writing and client comms
- Everyday coding
- Prompt and agent building
Will happily go long. Ask for tighter output.
The default general assistant with the widest tool and connector ecosystem.
- Research with browsing
- Data analysis in chat
- Mixed everyday tasks
Personality drifts between updates, so saved prompts can degrade.
Very fast, very cheap, and surprisingly capable on structured work.
- High volume summarisation
- Automation steps
- Classification
Loses nuance on ambiguous instructions.
Solid European option with strong multilingual output and EU hosting.
- EU data residency requirements
- Multilingual drafting
Behind the leaders on hard reasoning.
Small, quick, and cheap enough to run constantly in the background.
- Sub-agents
- Batch transcript processing
- Audit logging
Do not hand it work that needs judgement.
Cheap workhorse for high-frequency API calls inside automations.
- Tagging and routing
- Short extractions
- Webhook handlers
Needs tight, example-heavy prompts.
About as cheap as inference gets while still being usable.
- Bulk cleanup
- Simple rewrites at scale
Quality drops sharply on anything nuanced.
The most controllable image model for text-in-image and layout work.
- Thumbnails
- Social graphics
- Mockups with legible text
Style consistency across a series takes reference images.
Image to video that keeps the product recognisable while adding motion.
- Short vertical ad clips
- Product motion from one still
Short clip lengths, so plan the cut before you generate.
Still the benchmark for natural voice and real-time voice agents.
- Voice agents
- Narration
- Character voices
Costs scale with usage, so cap your agent minutes.
Full songs and production music from a prompt, fast enough to iterate live.
- Podcast themes
- Background beds
- Jingles
Check licensing terms before commercial use.