Skip to content

The frontier

As of November 16, 2026, the highest Analyzer Index on the price frontier is Gemini 4 Carbon at 95.7, with a blended price of $2.00 per million tokens. A model is on the frontier when no other model beats it on both capability and price.

Every dot is a model. The line joins models that no other model beats on both axes at once. Change the axes to trade capability against price, speed, latency or context; the URL keeps your view so you can share it. Capability is the Analyzer Index v1.0.

Where the frontier splits

Rank on each scenario for the models on this chart. A column’s #1 is the leaderboard leader. The middle card picks the highest quality score on each scenario among models with coverage close to the all-rounder’s, so a score from one slice of the benchmarks cannot win. Both figures are a benchmark-level estimate: one published score per scenario, with no live agent trace.

One model for every scenario

Claude Fable 5.1

Average quality 84.1 across 11 scenarios. Equal-weight monthly cost $4,005.68.

Highest quality on each scenario

7 models

Average quality 94.9, 10.8 points above Claude Fable 5.1. That mix averages $2,127.41 per month, about 47% less than Claude Fable 5.1.

Each scenario’s current leader

8 models

Average quality 89.0, 4.9 points above Claude Fable 5.1. These ranks already include each scenario’s cost weight. That mix averages $222.88 per month, about 94% less than Claude Fable 5.1.

Traffic mix

Each scenario has an equal share. Move a slider to weight the monthly cost.

Scenario ranks for models that lead at least one scenario. Rank 1 matches the current leaderboard leader for that scenario.
Model👨‍💻 Coding Assistant💬 Customer Support Bot🔬 Research / Analyst🧮 Math / Reasoning Tutor🏷️ Bulk Classification🤖 Autonomous Agent / Tool Use✍️ Writing & Content📚 Knowledge Q&A Assistant📊 Data / BI Analyst⚡ Realtime Chat / Voice🛡️ Production Critical
Claude Fable 5.1Best singleAnthropic27, rank 27 on Coding Assistant127, rank 127 on Customer Support Bot8, rank 8 on Research / Analyst16, rank 16 on Math / Reasoning Tutor65, rank 65 on Bulk Classification31, rank 31 on Autonomous Agent / Tool Use20, rank 20 on Writing & Content25, rank 25 on Knowledge Q&A Assistant30, rank 30 on Data / BI Analyst199, rank 199 on Realtime Chat / Voice17, rank 17 on Production Critical
Claude Opus 5.5Anthropic6, rank 6 on Coding Assistant125, rank 125 on Customer Support Bot6, rank 6 on Research / Analyst7, rank 7 on Math / Reasoning Tutor53, rank 53 on Bulk Classification9, rank 9 on Autonomous Agent / Tool Use17, rank 17 on Writing & Content15, rank 15 on Knowledge Q&A Assistant21, rank 21 on Data / BI Analyst186, rank 186 on Realtime Chat / Voice8, rank 8 on Production Critical
Claude Opus 4.7Anthropic36, rank 36 on Coding Assistant86, rank 86 on Customer Support Bot45, rank 45 on Research / Analyst43, rank 43 on Math / Reasoning Tutor62, rank 62 on Bulk Classification28, rank 28 on Autonomous Agent / Tool Use63, rank 63 on Writing & Content58, rank 58 on Knowledge Q&A Assistant33, rank 33 on Data / BI Analyst200, rank 200 on Realtime Chat / Voice4, rank 4 on Production Critical
Claude Fable 5.2Anthropic—, not ranked on Coding Assistant—, not ranked on Customer Support Bot1, rank 1 on Research / Analyst4, rank 4 on Math / Reasoning Tutor—, not ranked on Bulk Classification29, rank 29 on Autonomous Agent / Tool Use—, not ranked on Writing & Content3, rank 3 on Knowledge Q&A Assistant—, not ranked on Data / BI Analyst223, rank 223 on Realtime Chat / Voice1, rank 1 on Production Critical
Gemini 4 CarbonGoogle—, not ranked on Coding Assistant6, rank 6 on Customer Support Bot2, rank 2 on Research / Analyst1, rank 1 on Math / Reasoning Tutor—, not ranked on Bulk Classification1, rank 1 on Autonomous Agent / Tool Use1, rank 1 on Writing & Content4, rank 4 on Knowledge Q&A Assistant—, not ranked on Data / BI Analyst143, rank 143 on Realtime Chat / Voice—, not ranked on Production Critical
GPT-6 AstraOpenAI32, rank 32 on Coding Assistant139, rank 139 on Customer Support Bot3, rank 3 on Research / Analyst9, rank 9 on Math / Reasoning Tutor57, rank 57 on Bulk Classification6, rank 6 on Autonomous Agent / Tool Use55, rank 55 on Writing & Content20, rank 20 on Knowledge Q&A Assistant9, rank 9 on Data / BI Analyst207, rank 207 on Realtime Chat / Voice12, rank 12 on Production Critical
GPT-5.6 Luna ProOpenAI—, not ranked on Coding Assistant—, not ranked on Customer Support Bot—, not ranked on Research / Analyst—, not ranked on Math / Reasoning Tutor—, not ranked on Bulk Classification—, not ranked on Autonomous Agent / Tool Use—, not ranked on Writing & Content—, not ranked on Knowledge Q&A Assistant—, not ranked on Data / BI Analyst1, rank 1 on Realtime Chat / Voice—, not ranked on Production Critical
GLM 5.3 FlashZhipu AI (GLM)—, not ranked on Coding Assistant1, rank 1 on Customer Support Bot—, not ranked on Research / Analyst35, rank 35 on Math / Reasoning Tutor7, rank 7 on Bulk Classification35, rank 35 on Autonomous Agent / Tool Use14, rank 14 on Writing & Content23, rank 23 on Knowledge Q&A Assistant8, rank 8 on Data / BI Analyst23, rank 23 on Realtime Chat / Voice—, not ranked on Production Critical
GPT-6 LunaOpenAI15, rank 15 on Coding Assistant3, rank 3 on Customer Support Bot33, rank 33 on Research / Analyst13, rank 13 on Math / Reasoning Tutor1, rank 1 on Bulk Classification16, rank 16 on Autonomous Agent / Tool Use56, rank 56 on Writing & Content32, rank 32 on Knowledge Q&A Assistant14, rank 14 on Data / BI Analyst19, rank 19 on Realtime Chat / Voice2, rank 2 on Production Critical
Gemini 3 FlashGoogle1, rank 1 on Coding Assistant12, rank 12 on Customer Support Bot25, rank 25 on Research / Analyst55, rank 55 on Math / Reasoning Tutor—, not ranked on Bulk Classification25, rank 25 on Autonomous Agent / Tool Use13, rank 13 on Writing & Content12, rank 12 on Knowledge Q&A Assistant—, not ranked on Data / BI Analyst20, rank 20 on Realtime Chat / Voice—, not ranked on Production Critical
Muse Spark 1.3Meta—, not ranked on Coding Assistant28, rank 28 on Customer Support Bot—, not ranked on Research / Analyst24, rank 24 on Math / Reasoning Tutor20, rank 20 on Bulk Classification—, not ranked on Autonomous Agent / Tool Use6, rank 6 on Writing & Content1, rank 1 on Knowledge Q&A Assistant11, rank 11 on Data / BI Analyst100, rank 100 on Realtime Chat / Voice—, not ranked on Production Critical
DeepSeek V4 Flash 0731DeepSeek—, not ranked on Coding Assistant—, not ranked on Customer Support Bot51, rank 51 on Research / Analyst27, rank 27 on Math / Reasoning Tutor2, rank 2 on Bulk Classification32, rank 32 on Autonomous Agent / Tool Use110, rank 110 on Writing & Content45, rank 45 on Knowledge Q&A Assistant1, rank 1 on Data / BI Analyst10, rank 10 on Realtime Chat / Voice—, not ranked on Production Critical
  • Rank 1
  • Top 3
  • Top 10

This table is the shared model pool. Local / Self-Hosted Coding ranks a narrower set.

On the price frontier today

Models on the Pareto frontier, highest score first
ModelIndexBlended $/M
Gemini 4 CarbonGoogle95.7$2.000
Gemini 3.7 FlashGoogle82.4$0.750
DeepSeek V4.1 FlashDeepSeek78.5$0.210
DeepSeek V4 Flash 0731DeepSeek64.5$0.090
GPT-5 nanoOpenAI34.0$0.069

Trends over time

Frontier by scenario