AI Model Ranking
Snapshot: 2026-09-16 · Last reviewed: 2026-09-16
Metrics follow a snapshot of the Artificial Analysis public LLM leaderboard: Intelligence Index, cost per task, and speed reflect benchmark performance at the snapshot date. Provider pricing, availability, and rankings change frequently, and a high benchmark score does not guarantee the best result for a specific codebase or workflow.
Current reasoning model
Claude Fable 5.1 max still leads this snapshot at Intelligence Index 53 with 1M context. First-chunk latency remains very high; AA scores are not comparable 1:1 with the previous snapshot.
Best for: Highest-effort Claude Fable work when Intelligence Index matters more than wait time.
Anthropic
Current reasoning model
Claude Fable 5.1 xhigh rises to #2 at Intelligence Index 53 with 1M context. Cost per task is now public; first-chunk wait is much shorter than Fable 5.1 max.
Best for: High-effort Fable 5.1 coding when max first-chunk wait is too high and you still want near-top Intelligence Index.
Anthropic
Current reasoning model
GPT-6 Astra max is #3 with Intelligence Index 53 and 1M context. First-chunk wait is the longest in the top 10.
Best for: OpenAI-centered frontier agents that can absorb very long first-chunk delay.
OpenAI
Current reasoning model
GPT-6 Astra xhigh sits just below max with Intelligence Index 53, lower cost per task, and still-long first-chunk latency.
Best for: OpenAI coding agents that need near-max Astra quality at a lower task cost than max.
OpenAI
Current reasoning model
Claude Fable 5.1 high jumps into the top 5 at Intelligence Index 51 with 1M context. First-chunk latency is much lower than Fable max/xhigh.
Best for: Claude Fable 5.1 coding that still needs a high Intelligence Index without the max-tier first-chunk wait.
Anthropic
Current reasoning model
GPT-6 Astra high holds Intelligence Index 51 with 1M context; first chunk is much shorter than Astra max/xhigh.
Best for: OpenAI Astra agents that need strong reasoning without the max-tier first-chunk wait.
OpenAI
Current reasoning model
Claude Opus 5 max drops to #7. Intelligence Index 51 with 1M context; first chunk is shorter than Fable 5.1 max and Astra max.
Best for: Frontier Claude repository analysis and high-value agents when Fable 5.1 max wait is too high.
Anthropic
Current reasoning model
Claude Fable 5 stays in the top 10 at Intelligence Index 50 with 1M context; first-chunk wait remains long and cost per task is the highest in the top 10.
Best for: Deep repository analysis and high-value Claude Fable agent tasks where latency is acceptable.
Anthropic
Current reasoning model
GPT-6 Astra medium enters the top 10 at Intelligence Index 50 with 1M context and a 4.02s first chunk, much faster to first token than Astra max/xhigh.
Best for: Daily OpenAI Astra coding with 1M context when max/xhigh first-chunk wait is too high.
OpenAI
Current reasoning model
Opus 5 xhigh keeps Intelligence Index 50 with lower first-chunk latency than Opus 5 max and a lower cost per task.
Best for: Deep multi-file edits and coding agents that still need frontier quality with better cost than max.
Anthropic
Current reasoning model
Anthropic Claude Fable 5.1 (medium) ranks in this snapshot with Intelligence Index 49, 1M context, cost per task $2.98, median 51 tok/s, and first chunk 7.47s.
Best for: Anthropic coding agents, long-context work, and snapshot-based model comparison.
Anthropic
Current reasoning model
Anthropic Claude Opus 5 (high) ranks in this snapshot with Intelligence Index 48, 1M context, cost per task $3.61, median 49 tok/s, and first chunk 8.58s.
Best for: Daily Claude agent coding, review, and tool-heavy workflows that still need strong reasoning.
Anthropic
Current reasoning model
Meta Muse Spark 1.3 (max) ranks in this snapshot with Intelligence Index 48, 1M context, cost per task $1.60, median 221 tok/s, and first chunk 20.57s.
Best for: Fast Meta-stack coding with 1M context when you want high Intelligence Index and much lower wait than Fable/Astra max.
Meta
Current reasoning model
OpenAI GPT Sol 5.6 (max) ranks in this snapshot with Intelligence Index 47, 1M context, cost per task $1.99, median 65 tok/s, and first chunk 99.40s.
Best for: OpenAI coding agents, long-context work, and snapshot-based model comparison.
OpenAI
Current reasoning model
Anthropic Claude Fable 5.1 (low) ranks in this snapshot with Intelligence Index 47, 1M context, cost per task $2.37, median 50 tok/s, and first chunk 5.13s.
Best for: Anthropic coding agents, long-context work, and snapshot-based model comparison.
Anthropic
Current reasoning model
OpenAI GPT Astra 6 (low) ranks in this snapshot with Intelligence Index 46, 1M context, cost per task $0.82, median 51 tok/s, and first chunk 2.47s.
Best for: OpenAI coding agents, long-context work, and snapshot-based model comparison.
OpenAI
Current reasoning model
Alibaba Qwen 3.8 Max (0902) ranks in this snapshot with Intelligence Index 45, 984k context, cost per task $5.41, median 41 tok/s, and first chunk 2.72s.
Best for: Qwen-centered coding agents that want the 0902 Max snapshot rather than the undated Max SKU.
Alibaba
Current reasoning model
Meta Muse Spark 1.3 (xhigh) ranks in this snapshot with Intelligence Index 45, 1M context, cost per task $1.37, median 232 tok/s, and first chunk 22.12s.
Best for: Meta coding agents, long-context work, and snapshot-based model comparison.
Meta
Current reasoning model
Anthropic Claude Opus 5 (medium) ranks in this snapshot with Intelligence Index 45, 1M context, cost per task $2.19, median 48 tok/s, and first chunk 5.24s.
Best for: Anthropic coding agents, long-context work, and snapshot-based model comparison.
Anthropic
Current reasoning model
Z AI GLM 5.3 (max) ranks in this snapshot with Intelligence Index 45, 1M context, cost per task $2.01, median 67 tok/s, and first chunk 3.09s.
Best for: Z AI coding agents, long-context work, and snapshot-based model comparison.
Z AI
Current reasoning model
SpaceXAI Grok 4.6 (high) ranks in this snapshot with Intelligence Index 44, 500k context, cost per task $1.86, median 59 tok/s, and first chunk 36.19s.
Best for: SpaceXAI coding agents, long-context work, and snapshot-based model comparison.
SpaceXAI
Current reasoning model
SpaceXAI Grok 4.6 (xhigh) ranks in this snapshot with Intelligence Index 44, 500k context, cost per task $2.32, median 55 tok/s, and first chunk 33.14s.
Best for: SpaceXAI coding agents, long-context work, and snapshot-based model comparison.
SpaceXAI
Current reasoning model
OpenAI GPT Sol 5.6 (xhigh) ranks in this snapshot with Intelligence Index 44, 1M context, cost per task $1.18, median 57 tok/s, and first chunk 33.29s.
Best for: OpenAI coding agents, long-context work, and snapshot-based model comparison.
OpenAI
Current reasoning model
Kimi K3 (max) ranks in this snapshot with Intelligence Index 44, 1.05M context, cost per task $2.00, median 35 tok/s, and first chunk 4.52s.
Best for: Long-horizon coding agents, frontend generation, regional API stacks, and cost-aware 1M context work.
Kimi
Current reasoning model
SpaceXAI Grok 4.6 (medium) ranks in this snapshot with Intelligence Index 43, 500k context, cost per task $1.50, median 54 tok/s, and first chunk 32.75s.
Best for: SpaceXAI coding agents, long-context work, and snapshot-based model comparison.
SpaceXAI
Current reasoning model
OpenAI GPT Sol 5.6 (high) ranks in this snapshot with Intelligence Index 42, 1M context, cost per task $0.81, median 55 tok/s, and first chunk 9.34s.
Best for: OpenAI coding agents, long-context work, and snapshot-based model comparison.
OpenAI
Current reasoning model
OpenAI GPT Terra 5.6 (max) ranks in this snapshot with Intelligence Index 42, 1M context, cost per task $1.40, median 101 tok/s, and first chunk 139.76s.
Best for: OpenAI coding agents, long-context work, and snapshot-based model comparison.
OpenAI
Current reasoning model
Z AI GLM 5.3 Flash ranks in this snapshot with Intelligence Index 42, 1M context, cost per task $0.25, median 114 tok/s, and first chunk 2.45s.
Best for: Cost-sensitive GLM coding with 1M context and low first-chunk latency.
Z AI
Current reasoning model
Google Gemini 3.8 Flash (high) ranks in this snapshot with Intelligence Index 41, 1M context, cost per task $1.24, median 336 tok/s, and first chunk 16.28s.
Best for: Fast Google-stack coding loops, long-file skim, and high-throughput assistant work.
Current reasoning model
Alibaba Qwen 3.8 Max ranks in this snapshot with Intelligence Index 40, 1M context, cost per task $2.67, median 41 tok/s, and first chunk 2.59s.
Best for: Multilingual coding, regional API stacks, and Qwen-centered agent experiments.
Alibaba
Current reasoning model
Alibaba Qwen 3.8 2.4T A95B ranks in this snapshot with Intelligence Index 40, 984k context, cost per task $2.16, median 41 tok/s, and first chunk 2.77s.
Best for: Alibaba coding agents, long-context work, and snapshot-based model comparison.
Alibaba
Current model; partial public metrics
Google Gemini 3.8 Flash (medium) ranks in this snapshot with Intelligence Index 40, 1M context, cost per task $0.93, speed not fully public, and latency not fully public.
Best for: Google coding agents, long-context work, and snapshot-based model comparison.
Current reasoning model
Alibaba Qwen 3.8 Flash Next ranks in this snapshot with Intelligence Index 40, 256k context, cost per task $0.37, median 52 tok/s, and first chunk 2.67s.
Best for: Alibaba coding agents, long-context work, and snapshot-based model comparison.
Alibaba
Current reasoning model
Meta Muse Spark 1.2 (xhigh) ranks in this snapshot with Intelligence Index 40, 1.05M context, cost per task $0.97, median 181 tok/s, and first chunk 16.46s.
Best for: Meta coding agents, long-context work, and snapshot-based model comparison.
Meta
Current reasoning model
Anthropic Claude Opus 5 (low) ranks in this snapshot with Intelligence Index 40, 1M context, cost per task $1.10, median 46 tok/s, and first chunk 2.52s.
Best for: Anthropic coding agents, long-context work, and snapshot-based model comparison.
Anthropic
Current model; partial public metrics
Google Gemini 3.7 Flash (medium) ranks in this snapshot with Intelligence Index 40*, 1M context, cost per task is not fully public, median 293 tok/s, and first chunk 4.39s. Estimated Intelligence Index.
Best for: Google coding agents, long-context work, and snapshot-based model comparison.
Current reasoning model
DeepSeek V4.1 Flash (max) ranks in this snapshot with Intelligence Index 40, 1M context, cost per task $0.27, median 212 tok/s, and first chunk 1.31s.
Best for: Low-cost DeepSeek coding with 1M context, fast first chunk, and high output speed.
DeepSeek
Current reasoning model
OpenAI GPT Sol 5.6 (medium) ranks in this snapshot with Intelligence Index 39, 1M context, cost per task $0.50, median 58 tok/s, and first chunk 3.21s.
Best for: OpenAI coding agents, long-context work, and snapshot-based model comparison.
OpenAI
Current reasoning model
Google Gemini 3.7 Flash (high) ranks in this snapshot with Intelligence Index 39, 1M context, cost per task $0.93, median 292 tok/s, and first chunk 10.22s.
Best for: Google coding agents, long-context work, and snapshot-based model comparison.
Current reasoning model
SpaceXAI Grok 4.5 (high) ranks in this snapshot with Intelligence Index 39, 500k context, cost per task $1.04, median 60 tok/s, and first chunk 6.45s.
Best for: SpaceXAI coding agents, long-context work, and snapshot-based model comparison.
SpaceXAI
Current reasoning model
Anthropic Claude Sonnet 5 (max) ranks in this snapshot with Intelligence Index 38, 1M context, cost per task $5.09, median 84 tok/s, and first chunk 195.83s.
Best for: Anthropic coding agents, long-context work, and snapshot-based model comparison.
Anthropic
Current reasoning model
OpenAI GPT Terra 5.6 (xhigh) ranks in this snapshot with Intelligence Index 38, 1M context, cost per task $0.63, median 96 tok/s, and first chunk 14.46s.
Best for: OpenAI coding agents, long-context work, and snapshot-based model comparison.
OpenAI
Current reasoning model
OpenAI GPT Luna 5.6 (max) ranks in this snapshot with Intelligence Index 38, 1M context, cost per task $0.18, median 117 tok/s, and first chunk 124.21s.
Best for: OpenAI coding agents, long-context work, and snapshot-based model comparison.
OpenAI
Current model; partial public metrics
Google Gemini 3.7 Flash (low) ranks in this snapshot with Intelligence Index 37*, 1M context, cost per task is not fully public, median 308 tok/s, and first chunk 0.84s. Estimated Intelligence Index.
Best for: Very fast Google Flash loops when estimated Intelligence Index is enough and cost is still being measured.
Current reasoning model
DeepSeek V4 Pro 0813 (max) ranks in this snapshot with Intelligence Index 36, 1M context, cost per task $0.67, median 94 tok/s, and first chunk 1.72s.
Best for: DeepSeek coding agents, long-context work, and snapshot-based model comparison.
DeepSeek
Current model; partial public metrics
Sapiens AI Agnes 3.0 Flash ranks in this snapshot with Intelligence Index 36*, 1M context, cost per task is not fully public, median 238 tok/s, and first chunk 1.84s. Estimated Intelligence Index.
Best for: Fast Agnes coding when public cost is still incomplete and you need a 1M-context Flash SKU.
Sapiens AI
Current reasoning model
SpaceXAI Grok 4.6 (low) ranks in this snapshot with Intelligence Index 35, 500k context, cost per task $0.48, median 52 tok/s, and first chunk 5.63s.
Best for: SpaceXAI coding agents, long-context work, and snapshot-based model comparison.
SpaceXAI
Current model; partial public metrics
Sapiens AI Agnes 2.5 Pro Beta ranks in this snapshot with Intelligence Index 35*, 1M context, cost per task is not fully public, speed not fully public, and latency not fully public. Estimated Intelligence Index.
Best for: Experimental Agnes Pro Beta coding when only a partial public metric snapshot is available.
Sapiens AI
Current reasoning model
DeepSeek V4 Flash Vision (max) ranks in this snapshot with Intelligence Index 35, 1M context, cost per task $0.31, median 216 tok/s, and first chunk 1.16s.
Best for: DeepSeek coding agents, long-context work, and snapshot-based model comparison.
DeepSeek
Current reasoning model
OpenAI GPT Luna 5.6 (xhigh) ranks in this snapshot with Intelligence Index 35, 1M context, cost per task $0.09, median 117 tok/s, and first chunk 45.03s.
Best for: OpenAI coding agents, long-context work, and snapshot-based model comparison.
OpenAI