Which model should you pick?
Every AI tool now hands you a model menu. The names are noise. This page sorts the models I actually use by the job you are trying to do, and tells you where you will find each one.
Want to see the differences for yourself? Six models built the same game: play all six →
Direct
The architect tier: plan the work, design the approach, then direct cheaper models to do it, picking the right one for each piece. Right now Claude Opus does this job for almost everything; Fable only adds a little edge on very big, long-running tasks. I expect that to flip again with the next Fable release.
3 models
Claude Opus
Almost everything: planning, deep reasoning, writing, research, and complex agentic coding. Fable only adds an edge on very big, long-running tasks.
- Context
- 1M tokens
- Price
- $4 / $20
- Speed
- Moderate
Claude Fable
Very big, long-running agentic tasks where a little extra edge over Opus 5.5 is worth two and a half times the price.
- Context
- 1M tokens
- Price
- $10 / $50
- Speed
- Slow
GPT-6 Astra
Work where the answer has to be right the first time and the token bill is not the constraint.
- Context
- 1.05M tokens
- Price
- $10 / $50
- Speed
- Moderate
| Model | Best for | Context | Pricein / out per M tokens | Speed | Where | Rating |
|---|---|---|---|---|---|---|
| Claude Opus Fabian's pick Anthropic | Almost everything: planning, deep reasoning, writing, research, and complex agentic coding. Fable only adds an edge on very big, long-running tasks. | 1M tokens | $4 / $20 | Moderate | Claude, Claude Code, Claude Cowork, Perplexity Computer, Poe | |
| Claude Fable Anthropic | Very big, long-running agentic tasks where a little extra edge over Opus 5.5 is worth two and a half times the price. | 1M tokens | $10 / $50 | Slow | Claude, Poe, Claude Code | |
| GPT-6 Astra OpenAI | Work where the answer has to be right the first time and the token bill is not the constraint. | 1.05M tokens | $10 / $50 | Moderate | ChatGPT, OpenAI Codex, Poe, OpenRouter |
Think
Strong reasoning for hard problems: strategy, analysis, the calls where the quality of the thinking decides everything downstream. Claude Opus is my default here and for almost everything else. GPT-6 belongs here at its Sol tier.
3 models · Why "think expensive, execute cheap" →
Claude Opus
Almost everything: planning, deep reasoning, writing, research, and complex agentic coding. Fable only adds an edge on very big, long-running tasks.
- Context
- 1M tokens
- Price
- $4 / $20
- Speed
- Moderate
Gemini Argon
Precise technical work where you want instructions followed to the letter, like coding and analysis.
- Context
- 1M tokens
- Price
- $2 / $10
- Speed
- Moderate
GPT-6
Tiers: Sol · Luna
Execution work at scale: Sol when the task is hard or mistakes are expensive, Luna for high-volume work where speed and price matter most.
- Context
- 1.05M tokens
- Price
- Sol $2 / $10 · Luna $0.10 / $0.50
- Speed
- Fast
| Model | Best for | Context | Pricein / out per M tokens | Speed | Where | Rating |
|---|---|---|---|---|---|---|
| Claude Opus Fabian's pick Anthropic | Almost everything: planning, deep reasoning, writing, research, and complex agentic coding. Fable only adds an edge on very big, long-running tasks. | 1M tokens | $4 / $20 | Moderate | Claude, Claude Code, Claude Cowork, Perplexity Computer, Poe | |
| Gemini Argon Google | Precise technical work where you want instructions followed to the letter, like coding and analysis. | 1M tokens | $2 / $10 | Moderate | Gemini, Google AI Studio, Antigravity | |
| GPT-6 OpenAI · Sol · Luna | Execution work at scale: Sol when the task is hard or mistakes are expensive, Luna for high-volume work where speed and price matter most. | 1.05M tokens | Sol $2 / $10 · Luna $0.10 / $0.50 | Fast | ChatGPT, OpenAI Codex, Poe, OpenRouter |
Execute
The workhorses. Drafting, coding, formatting, summarising, agents that run all day. I rarely pick these by hand anymore: they do their best work as subagents, when a stronger model hands them a focused job. GPT-6 comes in two tiers here, Sol for the harder work and Luna for volume.
8 models
Claude Sonnet
Well-scoped everyday jobs, bug fixes, polished documents and slides, and the focused jobs a stronger model hands to a subagent.
- Context
- 1M tokens
- Price
- $2 / $10
- Speed
- Fast
Claude Haiku
Subagents and automations: summarizing, translating, extracting, rewriting, and simple scripting at high volume.
- Context
- 1M tokens
- Price
- $0.10 / $0.50
- Speed
- Instant
GPT-6
Tiers: Sol · Luna
Execution work at scale: Sol when the task is hard or mistakes are expensive, Luna for high-volume work where speed and price matter most.
- Context
- 1.05M tokens
- Price
- Sol $2 / $10 · Luna $0.10 / $0.50
- Speed
- Fast
GLM-5.3-Flash
Running an assistant or agent that works all day on tool calls, where cost per task matters more than personality.
- Context
- 1M tokens
- Price
- $0.15 / $0.50
- Speed
- Fast
Gemini Flash
High-volume execution work: coding, agents, copy, and anything where speed and price beat peak intelligence.
- Context
- 1M tokens
- Price
- $0.75 / $3.75
- Speed
- Instant
Grok 4.7
Running inside Grok Bot, and reading what people on X are saying about a topic right now.
- Context
- 500K tokens
- Price
- $2 / $6
- Speed
- Moderate
DeepSeek V4.1 Flash
Agents that run all day, and any task you need to run hundreds of times without watching the bill.
- Context
- 1M tokens
- Price
- $0.30 / $1.20
- Speed
- Fast
Qwen3.8-Max
Coding, multi-tool agents and office tasks at mid-tier prices.
- Context
- 1M tokens
- Price
- $2 / $6
- Speed
- Moderate
| Model | Best for | Context | Pricein / out per M tokens | Speed | Where | Rating |
|---|---|---|---|---|---|---|
| Claude Sonnet Fabian's pick Anthropic | Well-scoped everyday jobs, bug fixes, polished documents and slides, and the focused jobs a stronger model hands to a subagent. | 1M tokens | $2 / $10 | Fast | Claude, Claude Code | |
| Claude Haiku Anthropic | Subagents and automations: summarizing, translating, extracting, rewriting, and simple scripting at high volume. | 1M tokens | $0.10 / $0.50 | Instant | Claude, Claude Code | |
| GPT-6 OpenAI · Sol · Luna | Execution work at scale: Sol when the task is hard or mistakes are expensive, Luna for high-volume work where speed and price matter most. | 1.05M tokens | Sol $2 / $10 · Luna $0.10 / $0.50 | Fast | ChatGPT, OpenAI Codex, Poe, OpenRouter | |
| GLM-5.3-Flash Zhipu AI (Z.ai) | Running an assistant or agent that works all day on tool calls, where cost per task matters more than personality. | 1M tokens | $0.15 / $0.50 | Fast | OpenClaw, OpenRouter | |
| Gemini Flash Google | High-volume execution work: coding, agents, copy, and anything where speed and price beat peak intelligence. | 1M tokens | $0.75 / $3.75 | Instant | Gemini, Poe, Google AI Studio, Antigravity | |
| Grok 4.7 xAI | Running inside Grok Bot, and reading what people on X are saying about a topic right now. | 500K tokens | $2 / $6 | Moderate | Grok Bot, Poe, OpenRouter | |
| DeepSeek V4.1 Flash DeepSeek | Agents that run all day, and any task you need to run hundreds of times without watching the bill. | 1M tokens | $0.30 / $1.20 | Fast | OpenClaw, OpenRouter | |
| Qwen3.8-Max Alibaba Cloud | Coding, multi-tool agents and office tasks at mid-tier prices. | 1M tokens | $2 / $6 | Moderate | OpenRouter |
Run privately on your own machine
Open-weights models small enough for a phone, a laptop or a workstation. Nothing leaves the building. Weaker than the cloud tier, and that is the trade.
2 models
Gemma 4
Short texts, translations and reading images on your own device, with nothing sent to the cloud.
- Runs on
- Phone to workstation
- Price
- Free, runs locally
- Speed
- Fast
GLM-5.3-Flash
Running an assistant or agent that works all day on tool calls, where cost per task matters more than personality.
- Runs on
- 1M tokens
- Price
- $0.15 / $0.50
- Speed
- Fast
| Model | Best for | Runs on | Price | Speed | Where | Rating |
|---|---|---|---|---|---|---|
| Gemma 4 Fabian's pick Google | Short texts, translations and reading images on your own device, with nothing sent to the cloud. | Phone to workstation | Free, runs locally | Fast | Hugging Face | |
| GLM-5.3-Flash Zhipu AI (Z.ai) | Running an assistant or agent that works all day on tool calls, where cost per task matters more than personality. | 1M tokens | $0.15 / $0.50 | Fast | OpenClaw, OpenRouter |
Make images
No context window here, and prices per image or per month. What matters is what each one is good at.
5 models
GPT-Image-2.5
Creative image work and UI design alike: conversational editing, precise inpainting, and text-heavy visuals that have to stay legible.
- Good at
- Editing, legible text
- Price
- $30 / M image tokens
- Speed
- Fast
Nano Banana
A second take on anything you'd make with GPT-Image: layouts, infographics, product shots, edits.
- Good at
- A second opinion
- Price
- ~$0.03 / image (1K)
- Speed
- Fast
FLUX 3 Image
Photorealistic images and photo editing, when you want frontier quality without waiting.
- Good at
- Photos, photo editing
- Price
- $0.04–0.61 / image
- Speed
- Fast
Midjourney
Concept art, stylized illustrations, and images where the look matters more than the layout.
- Good at
- Art, mood, style
- Price
- $10–120 / month
- Speed
- Moderate
Grok Imagine
People who use Grok or X a lot and want good images without switching tools.
- Good at
- If you use Grok
- Price
- $0.04 / image
- Speed
- Fast
| Model | Best for | Good at | Price | Speed | Where | Rating |
|---|---|---|---|---|---|---|
| GPT-Image-2.5 Fabian's pick OpenAI | Creative image work and UI design alike: conversational editing, precise inpainting, and text-heavy visuals that have to stay legible. | Editing, legible text | $30 / M image tokens | Fast | ChatGPT, Poe | |
| Nano Banana Google DeepMind | A second take on anything you'd make with GPT-Image: layouts, infographics, product shots, edits. | A second opinion | ~$0.03 / image (1K) | Fast | Gemini, Google AI Studio | |
| FLUX 3 Image Black Forest Labs | Photorealistic images and photo editing, when you want frontier quality without waiting. | Photos, photo editing | $0.04–0.61 / image | Fast | ||
| Midjourney Midjourney | Concept art, stylized illustrations, and images where the look matters more than the layout. | Art, mood, style | $10–120 / month | Moderate | Midjourney | |
| Grok Imagine xAI | People who use Grok or X a lot and want good images without switching tools. | If you use Grok | $0.04 / image | Fast | Grok Imagine |
Make video
Two entries so far, both priced per second of footage.
2 models
Veo 3.1
High-fidelity cinematic shots and realistic physics simulation.
- Good at
- Cinematic shots, physics
- Price
- $0.05–0.60 / sec
- Speed
- Moderate
Grok Imagine Video
Stylized animation and rapid video clip generation.
- Good at
- Stylized clips
- Price
- $0.08 / sec
- Speed
- Fast
| Model | Best for | Good at | Price | Speed | Where | Rating |
|---|---|---|---|---|---|---|
| Veo 3.1 Fabian's pick Google DeepMind | High-fidelity cinematic shots and realistic physics simulation. | Cinematic shots, physics | $0.05–0.60 / sec | Moderate | Veo 3.1 | |
| Grok Imagine Video xAI | Stylized animation and rapid video clip generation. | Stylized clips | $0.08 / sec | Fast | Grok Imagine |
Voice
Speech that carries emotion, and that you can direct line by line.
1 model
| Model | Best for | Good at | Price | Speed | Where | Rating |
|---|---|---|---|---|---|---|
| Eleven v4 Fabian's pick ElevenLabs | Voiceovers, audiobooks, emotional dialogue, and voice agents (v4 Turbo). | Expressive speech | $0.08 / 1K characters | Fast | ElevenLabs |
No model matches that
Try the provider name (Anthropic, Google) or a tool you use (ChatGPT, Poe).
Missing a model you rely on? Tell me and I will put it through its paces.