AITOPAI热报
IndustryMulti-sourceSelected

Intelligence Index thresholds now far cheaper per task, Artificial Analysis finds

官方账号Artificial Analysis@ArtificialAnlys
Post· 11 sources
View the post
Why we picked it

The price of intelligence is collapsing at every fixed quality level while top-end capability prices keep climbing, so buyers face a widening spread between commodity and frontier performance. Teams benchmarking models should match their task's required score to the cheapest threshold model instead of defaulting to the highest scorer. The data also names concrete price leaders at each level, such as Claude Haiku 5.5 (xhigh) at US$0.12/task for a score of 41.

Artificial Analysis reports that once models reach an Intelligence Index threshold, the cost of hitting that level falls quickly. Reaching a score of 30 now costs 86x less per task than when first achieved: OpenAI's GPT-6 Luna (high) hits 30 for US$0.03/task versus US$2.49/task for Anthropic's Claude Sonnet 4.6, which first reached 30 in February 2026. Reaching 20 is 123x cheaper than roughly 13 months ago, and reaching 40 is 33x cheaper than about 5 months ago. Meanwhile, the highest measured score rose from 21 to 58 over one year at about 11x the cost, with Claude Opus 5.5 costing US$5.98/task. Figures are based on measured v4.3.2 results and prices as of 20 Oct 2026.

Original · Artificial Analysis

Once Intelligence Index thresholds are reached, the cost of accessing that level of intelligence reduces fast. Reaching 30 on the Intelligence Index now costs 86x less per task than when first reached

Existing intelligence thresholds have become cheaper to reach, while the highest-scoring model costs more than the highest-scoring from one year ago.

Key takeaways:

➤ Reaching 20 is 123x cheaper than ~13 months ago; reaching 40 is 33x cheaper than ~5 months ago: measured models first reached these thresholds in September 2025 and May 2026.

➤ GPT-6 Luna (high) reaches 30 for $0.03/task: @OpenAI's model is 86x cheaper per task than @AnthropicAI's Claude Sonnet 4.6 (Adaptive Reasoning, Max Effort), which first reached 30 in February 2026 at $2.49/task.

➤ Claude Haiku 5.5 (xhigh) is now cheapest to reach 40: it scores 41 at $0.12/task, 5% below GPT-6.1 Sol (low).

➤ Reaching 50 is 18x cheaper than when first reached: GPT-6.1 Sol (high) reaches it for $0.32/task.

➤ The highest measured score has risen from 21 to 58 in a year, at ~11x the cost: Claude Opus 5.5 (Adaptive Reasoning, Max Effort, Default Fallback) costs $5.98/task.

As of October 20, 2026, using measured v4.3.2 results and current prices, including for historical models.

#Artificial Analysis#OpenAI#Anthropic#GPT-6#Claude

Keep reading

Recent stories with the same tag or section
ProductsUS vs ChinaSelected12:24
DeepSeek Search ranks 17th on Artificial Analysis Search Index

The headline price is low, but the benchmark shows the real cost lives in tokens: ~7k-token search results make DeepSeek's model cost the 2nd highest on the board at ~US$0.023 per task. Developers building search-augmented agents should weigh that against the weak BrowseComp score of 63%, which trails Perplexity variants and Octen at 85% to 87% on multi-hop tasks. Compared with alternatives like Exa and Firecrawl, DeepSeek wins on price, not on factual accuracy.

#DeepSeek#Search API#Artificial Analysis#agents
官方账号Artificial Analysis@ArtificialAnlysSource
ModelsUS vs ChinaSelected02:31
AA-PersonalityBench Ranks Model Personality and Communication Style

The non-obvious insight is that correctness alone no longer separates models; style metrics like Slop and Conviction now offer a measurable way to pick a model for tone-sensitive use-cases such as legal drafting or creative writing. Teams building user-facing products can use the eight indexes to shortlist models matching their brand voice. The data covers Chinese models like GLM-5.3, DeepSeek V4.1 Flash and Qwen3.8 27B alongside Western ones, making direct stylistic comparison across the two ecosystems possible.

#AA-PersonalityBench#Artificial Analysis#Claude#GLM
官方账号Artificial Analysis@ArtificialAnlysSource
ModelsBuildersMulti-sourceSelected19:14
Liquid AI releases d1-ultra, topping Artificial Analysis Decision Index

The trade-off here is deliberate: Liquid AI swapped raw accuracy for knowing when not to answer, which matters more for agentic decision workflows where a wrong answer is costlier than a refusal. Builders routing high-volume decisions can compare it directly against Celeris-1 Decision ($32 per 1M tasks), Mercury Decide ($15) and GPT-6 Luna (Decisions) ($79) on both price and speed.

#Liquid AI#d1-ultra#decision models#benchmarks
Event
官方账号Artificial Analysis@ArtificialAnlys3 sourcesSource
ProductsBuildersMulti-sourceSelected03:25
Artificial Analysis Adds Cache Hit Rates and Cost per Task

Cache pricing, not headline token rates, is now the dominant variable in agentic inference cost, so the cheapest sticker price can be up to 9x more expensive per task. Teams running long multi-turn coding agents can use these published hit rates and per-provider Cost per Task figures to pick endpoints and negotiate cache discounts. Z AI's first-party API hits 100% cache hit rate, a useful benchmark when comparing open-weights hosting options.

#GLM-5.3#Artificial Analysis#cache hit rate#open-source models
Event
官方账号Artificial Analysis@ArtificialAnlys4 sourcesSource
ProductsBuildersSelected00:45
Artificial Analysis Launches Decision Index v1.0 Benchmark

Decision models trade some accuracy for far lower cost and latency, making them viable for high-volume routing, moderation and classification workloads where a full language model is overkill. The finding that most models are poorly calibrated on knowledge tasks, yet reasonably calibrated on classification, tells builders where this approach is safe to deploy. Teams weighing Jev 1.13 against Microsoft-Decision-1 or Perplexity's Decider v1.1 get a direct cost-per-task comparison.

#Artificial Analysis#Jev#decision models#benchmarks
官方账号Artificial Analysis@ArtificialAnlysSource
ProductsBuildersSelected09:23
Artificial Analysis Video Leaderboards Update

This change addresses a critical need for transparency in AI model comparisons, as derived models often perform differently from their base models despite sharing architecture. For developers and businesses evaluating video generation options, this feature enables more informed decisions by allowing direct comparison of base models while still acknowledging distinct derived options.

#Artificial Analysis#AA-Video-T2V#MiniMax H3#Utopai X
Event
官方账号Artificial Analysis@ArtificialAnlys2 sourcesSource