The picks
For every kind of AI work: the best value for money, the best there is, the best with open weights where there is one, and the cheapest that still works. Each is a single ranking with a single stated rule, rebuilt every night from the boards and the price lists.
Text
Work done in words: the everyday model, the one that writes code, the one that can be sent off to finish a job, and the one that does the mathematics.
General
Which model to reach for when the job is not a specialised one. 195 measured, 160 of them for sale, and 62 of those stand on two or more of these boards — only those can be picked.
- Best value for money · Best frontierLlama 3.1 405BMeta · 1st of 137 on Wino grande · 3 boards$1.2→$1.2per Mtok in / out3 sellers
- Best open sourcePhi-3-medium-128k-instructMicrosoft · 5th of 134 on Arc ai2 · 2 boards$0.17→$0.68per Mtok in / out1 seller
- CheapestQwen2.5-Coder-7BAlibaba · 38th of 134 on Arc ai2 · 2 boards$0.025→$0.05per Mtok in / out2 sellers
Coding
Which model should write and fix code. 119 measured, 113 of them for sale, and 78 of those stand on two or more of these boards — only those can be picked.
- Best value for money · Best open sourceDeepSeek-R1 (0528)DeepSeek · 5th of 28 on LiveCodeBench Leaderboard · 2 boards$0.2→$0.25per Mtok in / out13 sellers
- Best frontierClaude Fable 5Anthropic · 1st of 30 on Frontiercode · 4 boards$10→$50per Mtok in / out15 sellers
- CheapestGLM 5.3 FlashZ.ai · 10th of 121 on Webdev arena · 3 boards$0.075→$0.25per Mtok in / out17 sellers
Agents
Which model can call tools and finish a job. 152 measured, 135 of them for sale, and 76 of those stand on two or more of these boards — only those can be picked.
- Best value for moneyMuse Spark 1.1Meta · 7th of 110 on OSWorld-Verified Results · 5 boards$1.25→$4.25per Mtok in / out4 sellers
- Best frontierGPT 6 AstraOpenAI · 2nd of 65 on Apex agents · 2 boards$10→$50per Mtok in / out15 sellers
- Best open sourceGLM 5.2Z.ai · 5th of 60 on Vending-Bench 2 · 4 boards$0.49→$1.73per Mtok in / out24 sellers
- CheapestDeepSeek V3.2 ExpDeepSeek · 1st of 16 on The agent company · 3 boards$0.21→$0.32per Mtok in / out8 sellers
Reasoning
Which model can do the mathematics and the science. 197 measured, 188 of them for sale, and 164 of those stand on two or more of these boards — only those can be picked.
- Best value for money · Best open sourceStep 3.5 FlashStepFun · 3rd of 61 on AIME 2025 · 2 boards$0.09→$0.3per Mtok in / out8 sellers
- Best frontierClaude Opus 5Anthropic · 1st of 23 on Gbaeval · 3 boards$5→$25per Mtok in / out15 sellers
- CheapestDeepSeek V4 Flash (0731)DeepSeek · 18th of 127 on Mystery game puzzles · 3 boards$0.04→$0.1per Mtok in / out18 sellers
Visual
Making a picture or a clip, and reading one.
Video generation
Which model makes the clip. 20 measured, 4 of them you can buy today.
Image generation
Which model draws the picture. 22 measured, 17 of them you can buy today.
- Best value for moneyMAI Image 2.5Microsoft · 5th of 153 on Artificial Analysis · Text to Image Arena$0.05an image1 seller
- Best frontierReve 2.1Reve · 2nd of 153 on Artificial Analysis · Text to Image Arena$0.2an image1 seller
- CheapestQwen-Image-3.0Alibaba · 11th of 153 on Artificial Analysis · Text to Image Arena$0.027an image1 seller
Image recognition
Which model understands a picture or a document. 56 measured, 55 of them you can buy today.
Voice
Speaking, listening, and holding a conversation aloud — three different markets that are easy to mistake for one.
Speech generation
Which model reads your text out loud. 31 measured, 12 of them you can buy today.
- Best value for moneyspeech-2.8-turboMiniMax · 14th of 39 on TTS Arena V2$60per million characters2 sellers
- Best frontierspeech-2.8-hdMiniMax · 9th of 39 on TTS Arena V2$100per million characters1 seller
- Cheapestinworld-tts-1.5-maxInworld AI · 15th of 39 on TTS Arena V2$50per million characters1 seller
Speech recognition
Which model turns speech into text. 25 measured, 8 of them you can buy today.
Live voice
Which model can take a call and handle it. 9 measured, 4 of them you can buy today.
Retrieval
The plumbing under a search box: turning text into vectors, and putting the right answer at the top.
Embeddings
Which model to build a search index on. 43 measured, 8 of them you can buy today.
- Best value for moneyjina-embeddings-v5-text-smallJina AI · 16th of 183 on MTEB · English · 2 boards$0.05per million tokens1 seller
- Best frontierGemini EmbeddingGoogle · 9th of 183 on MTEB · English · 2 boards$0.15per million tokens5 sellers
- Cheapestjina-embeddings-v5-text-nanoJina AI · 22nd of 183 on MTEB · English · 2 boards$0.05per million tokens1 seller
The market in halves
The same question asked of each side of the industry. Open weights are measured far less often than closed ones, so that block is the thinnest on the page, and it is last for that reason.
Open weights
Which model to run yourself, or buy from whoever you like. 203 measured, 132 of them for sale, and 89 of those stand on two or more of these boards — only those can be picked.
Proprietary
Which model behind an API is worth the money. 165 measured, 152 of them for sale, and 125 of those stand on two or more of these boards — only those can be picked.
- Best value for moneyo4 Mini HighOpenAI · 1st of 28 on LiveCodeBench Leaderboard · 2 boards$1.1→$4.4per Mtok in / out3 sellers
- Best frontierGPT 6 AstraOpenAI · 1st of 127 on Mystery game puzzles · 4 boards$10→$50per Mtok in / out15 sellers
- CheapestQwen3.7 FlashAlibaba · 69th of 222 on Chess puzzles · 2 boards$0.03→$0.13per Mtok in / out4 sellers
Every pick has to be something you can buy today: a model with no seller and no published price is not an answer to what should I use. Prices are the standard lane, never a batch or promotional rate. Figures read 2026-09-16.