AI tools
43 tools that are actually in use, grouped by job. Prices were public when written and can change — check the vendor site. Each entry says why it is worth using and when it is not.
Chat assistants
9The everyday entry point for Q&A, writing and analysis
ClaudeAnthropicFree tier
The steadiest option for long-document analysis and structured writing
Free tier runs Sonnet 5.5; Pro ~$20/mo; Max from ~$100/mo
The free tier already gives you Sonnet 5.5 (intelligence 54.2), one of the strongest models available for free anywhere. Artifacts and Projects make it easy to break long tasks into reusable workspaces.
ChatGPTOpenAIFree tier
The broadest surface: agents, images, voice and a plugin ecosystem
Free tier; Plus ~$20/mo; Pro ~$200/mo
The paid chat window runs GPT-6 Sol; GPT-6.1 Terra is only in Work and Codex. The most complete ecosystem, but consumer chats may be used for training by default — turn it off manually.
GeminiGoogleFree tier
1M context plus natively multimodal — first choice for research and long documents
Free tier; AI Pro ~$20/mo; AI Ultra ~$250/mo
AI Pro and Ultra users get Gemini 4 Argon from 13 October (GPQA 96.4, first overall). Deep Workspace integration is the main advantage.
GrokxAIFree tier
Real-time information with a creative bent
Capped free tier; SuperGrok ~$30/mo
Consistently strong on creativity evaluations, but 500K context limits long-document work.
KimiMoonshot AIFree tier
China's long-context specialist, with a 1M context window
Generous free tier; membership from ~¥99/mo
K3.1 Pro ranks first on Terminal-Bench 4.0, but 28 tok/s means long answers take a while. Built for cases where thinking comes before writing.
DeepSeekDeepSeekFree tier
The best value here — API pricing an order of magnitude lower
Web app free; API from $0.30/$1.20, half price off-peak
V4.1 Flash is MIT licensed with downloadable weights. The API has peak and off-peak pricing, so scheduling batch jobs overnight halves the cost.
QwenAlibabaFree tier
The most complete open ecosystem, with a 27B model you can run on one machine
Free; API billed by usage
The Qwen3.8 family runs from a 27B dense model to the closed Max, and the Apache-2.0 licence makes it the easiest to build on.
DoubaoByteDanceFree tier
Natural spoken Chinese and a good voice interface
Mostly free; enterprise edition priced separately
Low latency on real-time voice, a good fit for prototyping support and companion products.
Tencent YuanbaoTencentFree tier
Calls straight into the WeChat ecosystem
Free
Wired into WeChat Official Accounts and WeChat Read, which makes Chinese-source research easy.
Coding & development
9IDEs, CLI agents, completion and repo-wide refactoring
Claude CodeAnthropic
A repo-level agent in your terminal that stays on task
Included with a Claude subscription; API billed by usage
Reads the repo, runs tests and edits multiple files from the shell. Opus 5.5's Terminal-Bench 65.2% is essentially a score for this kind of tool.
CursorAnysphereFree tier
In-IDE completion and multi-file refactoring
Free tier; Pro ~$20/mo; Ultra ~$200/mo
Its tab-completion prediction accuracy is still the best in class, and you can switch between GPT, Claude and Gemini.
CodexOpenAI
Runs coding tasks asynchronously in the cloud
Included in paid ChatGPT tiers; API billed by usage
You file the task and it runs in the cloud — built for filing a batch of issues and letting it open the PRs.
GitHub CopilotGitHubFree tier
Tightest coupling to the PR and issue workflow
~$10/mo individual; ~$19/seat business
Copilot Review is what separates it from other IDE extensions.
WindsurfCognitionFree tier
Cascade multi-step editing
Free tier; Pro ~$15/mo
Its context index is lightweight, so large repositories load faster than rivals on first open.
ClineOpen weightsFree tier
An open-source agent inside VS Code, bring your own API key
Open source and free; you pay model API costs
No vendor lock-in — point it at a cheap API like DeepSeek and costs drop to 1/20.
AiderOpen weightsFree tier
Pair programming in the terminal, with automatic git commits
Open source, free
Every change gets a generated commit message, which suits teams that need a traceable history.
v0VercelFree tier
React and Tailwind interfaces from plain language
Free tier; Pro ~$20/mo
Output is shadcn/ui components you can paste straight in — good for testing a UI direction fast.
LovableLovableFree tier
Full-stack apps from a conversation
Free tier; Pro ~$25/mo
Supabase integration gets you to a login-capable MVP from nothing. Complex business logic still has to be written by hand.
Image generation & editing
5Text-to-image, image-to-image, inpainting and design work
ChatGPT ImagesOpenAI
The strongest text rendering and style consistency
Included with a ChatGPT subscription
The most reliable at Chinese and English typesetting in posters and UI mockups, and long text rarely breaks.
Nano BananaGoogleFree tier
Image-to-image with character-consistent edits
Included with Gemini; API from $1.50 per million tokens
Version 2.1 beats Grok Imagine on three of four evaluation boards, with character consistency its strong point.
MidjourneyMidjourney
The most distinctive aesthetic — first choice for art-led work
From ~$10/mo
Since v7 it balances realism and stylisation better, though text rendering is still weak.
FluxBlack Forest LabsFree tier
Open weights you can deploy and fine-tune locally
Open source; API billed per image
The most mature LoRA ecosystem — good for commercial work that needs a fixed character or art style.
ComfyUIOpen weightsFree tier
Node-based image and video workflows
Open source, free
Chains Flux, SD and video models into reusable pipelines; the de facto standard for local production.
Video generation
5Text-to-video, image-to-video and digital humans
VeoGoogle
Cinematic clips with native audio
Included with Gemini AI Pro/Ultra
Camera language and physical consistency are among the best here — good for advertising storyboards.
SoraOpenAI
Long-take storytelling and scene extension
Included in paid ChatGPT tiers
The storyboard tool lets you pace shot by shot, but generation still takes a long time.
KlingKuaishouFree tier
Top tier among Chinese video generators
Free credits; membership from ~¥66/mo
Natural human motion, good Chinese prompt comprehension, and direct access from mainland China.
HailuoMiniMaxFree tier
Camera movement and stability
Free credits; subscription tiers
H3 does 2K, 15 seconds with native audio — the longest of the Chinese options.
RunwayRunwayFree tier
A controllable toolchain for professional post-production
Free tier; Standard ~$15/mo
Motion brush and camera controls — manual intervention is where it beats pure generation models.
Speech & music
3TTS, voice cloning, music generation and transcription
ElevenLabsElevenLabsFree tier
Multilingual TTS and voice cloning
Free tier; Creator ~$22/mo
Chinese prosody is clearly more natural than most open-source TTS, with fine-grained emotional control.
SunoSunoFree tier
Whole songs, vocals included
Free tier; Pro ~$10/mo
Chinese enunciation became usable after v5 — fine for short-video soundtracks.
WhisperOpenAI / open sourceFree tier
The transcription baseline, runnable locally
Open source, free
Chinese punctuation and sentence breaks improved markedly after large-v3, and with VAD it handles long-audio pipelines.
Agents & automation
4Workflow orchestration, browser control and multi-step tasks
n8nn8nFree tier
Self-hosted visual automation that talks to any API
Self-hosting free; cloud from ~€20/mo
400+ nodes let you drop LLM calls into business processes, and self-hosting keeps data on your own network.
DifyLangGeniusFree tier
Open-source LLM app orchestration and RAG
Open source and free; the cloud version has a free tier
A polished Chinese UI covering prompts, knowledge base and API publishing in one chain — good for internal tools.
ManusManus
A general-purpose agent that opens a browser and gets on with it
From ~$19/mo
Built for research, price comparison and form filling — tasks with a clear endpoint but unclear steps.
Browser UseOpen weightsFree tier
Lets a model drive the browser directly
Open source, free
Abstracts the DOM into a clickable element tree — steadier than screenshot-only approaches, and good for automated testing.
Search & research
3Web search, paper retrieval and deep research
PerplexityPerplexityFree tier
Web answers with citations
Free tier; Pro ~$20/mo
Every claim carries a source link, which makes fact-checking easier than with a plain chat model.
NotebookLMGoogleFree tier
Turns a document set into a knowledge source you can question
Free tier; Pro included with AI Pro
After you upload PDFs or pages it answers only within them, with far fewer hallucinations than general models — good for reading papers.
ElicitElicitFree tier
Systematic literature review and data extraction
Free tier; Plus ~$12/mo
Pulls methods, sample sizes and conclusions out of hundreds of papers into a comparison table — a review-writing accelerator.
Inference & deployment
5Local inference, self-hosted serving and API routing
OllamaOpen weightsFree tier
Local models in one command
Open source, free
Runs Qwen3.8-27B, Gemma 4, gpt-oss and more, working out of the box on a Mac. For production, switch to vLLM.
LM StudioLM StudioFree tier
Graphical local model management
Free for personal use; commercial use needs a licence
Built-in model search and quantisation options make it the easy route for people who avoid the command line.
vLLMOpen weightsFree tier
A production-grade high-throughput inference engine
Open source, free
PagedAttention plus continuous batching typically gives several times Ollama's throughput on the same GPU. The de facto standard for self-hosted APIs.
OpenRouterOpenRouterFree tier
One API for every model, with price comparison built in
Usage-based pricing, no monthly fee
Compare quotes for the same model across providers and set automatic failover. The least painful way to evaluate several models.
Hugging FaceHugging FaceFree tier
Weights, datasets and inference endpoints
Free; Pro ~$9/mo
The most authoritative source for parameter counts, licences and context windows — a large part of this site's data comes from here.