Find the right AI for the job, and what it costs.

Claude Opus 4.5

Strongest of the Claude line for long reasoning and code. Priced accordingly — use Sonnet unless you actually need the extra depth.

Claude Sonnet 4.6

The sensible default: most of Opus's quality at a fraction of the cost.

GPT-5.5

Huge context window and strong general reasoning. Output tokens cost six times input, so verbose prompting gets expensive fast.

GPT-5 mini

The cheap tier for classification and cleanup work that doesn't need a frontier model.

Gemini 2.5 Pro

A million-token context window makes it the pick for whole-codebase or whole-document work.

Gemini 2.5 Flash

Built for speed and volume rather than depth. Good for bulk tasks, weaker at hard reasoning.

DeepSeek V4

Near-frontier quality at a fraction of Western API pricing, and the weights are public. Check where your data goes before putting client work through the hosted API.

Grok 4

Competitive on reasoning benchmarks. Smaller ecosystem and fewer integrations than the big three.

Mistral Large

EU-hosted, which matters if data residency is a hard requirement for you.

GPT Image 1

Best-in-class at following a detailed prompt and rendering legible text in an image.

DALL·E 3

Flat per-image pricing makes budgeting simple. Superseded by GPT Image 1 on quality.

FLUX.1 dev

The open-weights favourite — the most liked image model on Hugging Face. Non-commercial licence on the dev variant; read it before client work.

Stable Diffusion XL

Runs on consumer hardware and has the deepest ecosystem of LoRAs and fine-tunes. Older, and weaker than current models on prompt adherence.

Midjourney

Still the best pure aesthetic quality out of the box. No free tier, and no real API — it is a subscription product, not infrastructure.

Veo 3.1

Generates native audio alongside video. Billed per second of output, which adds up quickly on iteration.

Veo 2

Cheaper per second than 3.1 and fine for silent b-roll and motion tests.

Runway Gen-4

The most complete editing suite around the model, not just raw generation. Credit-based pricing is hard to predict month to month.

HunyuanVideo

The leading open-weights video model. Needs a serious GPU — this is not a laptop workload.

ElevenLabs v3

The quality benchmark for synthetic voice and cloning. Billed per character, so long-form narration costs real money.

Whisper

Transcription, not synthesis. Open weights mean you can run it locally for free instead of paying per minute.

Kokoro 82M

Tiny enough to run on a laptop and startlingly good for its size. Fewer voices and languages than the commercial options.

Suno

Full songs with vocals from a text prompt. Commercial rights only on paid tiers — check before using anything in client work.

Stable Audio Open

Open weights, trained on licensed audio, which makes provenance far less murky. Built for samples and loops rather than finished songs.

ACE-Step

Fast open-weights music generation you can self-host. Rougher output than Suno.

gpt-6-sol

gpt-6-luna

gpt-6-astra

gpt-6.1-sol

xai/grok-4.7

gpt-5.5-cyber

gpt-image-2.5-flare

gpt-image-2.5-sunburst

xai/grok-imagine-video

xai/grok-imagine-video-1.5

xai/grok-voice-transcribe-1.0

xai/grok-voice-transcribe-2.0

DeepSeek-R1

Llama-3.1-8B-Instruct

Meta-Llama-3-8B

DeepSeek-V4-Pro

gpt-oss-120b

Meta-Llama-3-8B-Instruct

GLM-5.2

gpt-oss-20b

stable-diffusion-xl-base-1.0

stable-diffusion-v1-4

FLUX.1-schnell

Z-Image-Turbo

stable-diffusion-3-medium

stable-diffusion-3.5-large

OrangeMixs

Sulphur-2-base

Wan2.1-T2V-14B

mochi-1-preview

MiniMax-H3-Turbo-Lora

HunyuanVideo-1.5

AnimateDiff-Lightning

pyramid-flow-sd3

XTTS-v2

Dia-1.6B

VibeVoice-1.5B

csm-1b

Qwen3-TTS-12Hz-1.7B-CustomVoice

chatterbox

VoxCPM2

stable-audio-open-1.0

ChatTTS

MiniMax-Music3

YuE2-3B

Ace-Step1.5

ACE-Step-v1-3.5B

riffusion-model-v1

musicgen-large