24 families, 135 models
Text and chat
Text models differ less by how well they write than by how far they can reason, how much they can read at once, whether they can use tools, and where your words go.
Our advice
Which one to take.
Or start from what you want to do
- Write, summarize and answer questions every dayGLMThe default: strong, private, works with every Sub Rosa tool.Or Claude
- Think through a hard problem or a long taskKimiHolds long chains of steps and tools, in private mode.Or GPT
- Write or fix codeClaudeCareful code and long agent sessions, anonymized.Or Kimi
- Ask about a photo, a screenshot or a scanned pageKimiReads images and keeps no data; Sub Rosa picks it for you.Or Gemini
- Work over a very long document or many notes at onceDeepSeekA very large context at a low price, in private mode.Or Gemini
- Get quick answers for very littleMercuryWrites many words at once: replies arrive almost instantly.Or Qwen open models
- Keep a sensitive conversation as private as possibleGLMEvery version is private, and some are listed as end-to-end encrypted.Or Kimi
- Write fiction or role play without refusalsVenice UncensoredBuilt to follow creative requests others decline, private.Or Uncensored community models
Quality and price
What you get for what you pay.
Higher is better. Each model is listed at its highest reasoning effort. Two or three points apart is a close call; ten points is a different class of model.
Each point is a family’s best measured version. The top left holds the best value.
See the numbers
| Model | Artificial Analysis Intelligence Index | US dollars per million tokens written (log scale) |
|---|---|---|
| Claude Opus 5.5 | 58 points | $8.992 |
| GPT-6 Astra | 53 points | $22.48 |
| Gemini 4 Argon | 53 points | $5.62 |
| Muse Wonder | 52 points | $11.24 |
| Grok 4.7 | 46 points | $3.057 |
| GLM 5.3 | 45 points | $2.473 |
| Qwen 3.8 Max | 45 points | $3.372 |
| Kimi K3 | 44 points | $8.43 |
| Qwen 3.8 2.4T | 40 points | $3.372 |
| DeepSeek V4.1 Flash | 39 points | $0.54 |
| MiMo-V2.6-Flash | 38 points | $0.157 |
| NVIDIA Nemotron 3.5 Super VL 120B | 26 points | $0.562 |
| Inkling | 25 points | $2.276 |
| MiniMax M2.7 | 23 points | $0.674 |
| Google Gemma 4 26B A4B Instruct | 17 points | $0.18 |
| Mercury 2 | 14 points | $0.422 |
| OpenAI GPT OSS 120B | 12 points | $0.135 |
| Llama 3.3 70B | 8 points | $1.259 |
| Mistral Small 3.2 24B Instruct | 8 points | $0.112 |
Leaderboards
Measured side by side.
One score averaging about ten independent tests of reasoning, knowledge, math, coding and agent work, all run the same way for every model.
See the numbers
| Model | points |
|---|---|
| 1. Claude Opus 5.5 | 58 |
| 2. Claude Sonnet 5.5 | 56 |
| 3. Claude Fable 5.1 | 53 |
| 4. GPT-6 Astra | 53 |
| 5. Gemini 4 Argon | 53 |
| 6. GPT-6.1 Sol | 52 |
| 7. Muse Wonder | 52 |
| 8. Claude Opus 5 | 51 |
| 9. Claude Fable 5 | 50 |
| 10. GPT-6 Sol | 48 |
Prices
Every family, side by side.
One row per family, at its recommended version, in US dollars. Sort any column; tick up to three families to compare them in detail.
| Compare | higher is better | per million tokens written | tokens | Privacy | |
|---|---|---|---|---|---|
| ClaudeClaude Sonnet 5.5 · Anthropic | 56 | $4.496 | 1,000K | Anonymized | |
| GeminiGemini 4 Argon · Google | 53 | $5.62 | 1,000K | Anonymized | |
| GPTGPT-6.1 Sol · OpenAI | 52 | $4.496 | 1,050K | Anonymized | |
| MuseMuse Wonder · Meta | 52 | $11.24 | 1,049K | Anonymized, some versions private | |
| GrokGrok 4.7 · xAI | 46 | $3.057 | 500K | Private | |
| Qwen Max, Plus and FlashQwen 3.8 Flash · Alibaba (Qwen) | 45Qwen 3.8 Max | $0.22 | 1,000K | Anonymized | |
| DeepSeekDeepSeek V4.1 Flash · DeepSeek | 39 | $0.54 | 1,000K | Private | |
| Xiaomi MiMoMiMo-V2.6-Flash · Xiaomi | 38 | $0.157 | 1,000K | Private | |
| GLMGLM 5.2 · Z.ai (Zhipu AI) | 34 | $1.978 | 1,000K | Private, some versions anonymized | |
| Qwen open modelsQwen 3.8 27B · Alibaba (Qwen) | 34 | $1.439 | 262K | Private | |
| KimiKimi K2.6 · Moonshot AI | 27 | $1.574 | 256K | Private | |
| NemotronNVIDIA Nemotron 3.5 Super VL 120B · NVIDIA | 26 | $0.562 | 262K | Private | |
| InklingInkling · Thinking Machines Lab | 25 | $2.276 | 524K | Private | |
| MiniMaxMiniMax M2.7 · MiniMax | 23 | $0.674 | 198K | Private | |
| GemmaGoogle Gemma 4 31B Instruct · Google DeepMind | 15 | $0.162 | 256K | Private | |
| GPT OSSOpenAI GPT OSS 120B · OpenAI | 12 | $0.135 | 128K | Private | |
| MercuryMercury 2.5 · Inception | 12 | $0.422 | 260K | Anonymized | |
| Llama and Hermes 3Llama 3.3 70B · Meta (Hermes 3: Nous Research) | 8 | $1.259 | 128K | Private | |
| Mistral SmallMistral Small 3.2 24B Instruct · Mistral AI | 8 | $0.112 | 256K | Private | |
| SeedSeed 2.1 Turbo · ByteDance | Not measured | $1.405 | 256K | Anonymized | |
| Venice UncensoredVenice Uncensored 1.2 · Venice.ai and Cognitive Computations (Dolphin) | Not measured | $0.405 | 128K | Private | |
| AionAion 3.5 · AionLabs | Not measured | $3.372 | 262K | Anonymized | |
| Uncensored community modelsGemma 4 Uncensored · Community fine-tunes (Abliteration AI, Olafangensan, others) | Not measured | $0.225 | 256K | Private, some versions anonymized | |
| HyHy4 Preview · Tencent | Not measured | $1.406 | 1,049K | Private |
Release history
The latest versions.
Prices read on October 16, 2026, in US dollars; scores dated on each chart. Sub Rosa shows the exact price before you start.