{"slug": "show-hn-independent-llm-benchmark-index-tracking-22-frontier-models", "title": "Show HN: Independent LLM benchmark index tracking 22 frontier models", "summary": "An independent LLM benchmark index launched on Hacker News' Show HN, ranking 22 frontier models on the Artificial Analysis Intelligence Index, coding proficiency, GPQA, SWE-bench Verified and HLE reasoning tests, plus real-time value-for-money ($/point) metrics updated daily. GPT-6 Astra (max) from OpenAI tops the intelligence index at 55, ahead of Anthropic's Claude Fable 5.1 (max with fallback) and Claude Opus 5 (max) at 53 each, Meta's Muse Spark 1.3 (max) and OpenAI's GPT-5.6 Sol (max) at 48, and Z AI's GLM-5.3 (max), Alibaba's Qwen3.8 Max (0902), StepFun's Step 5 Preview, SpaceXAI's Grok 4.6 (high) and Kimi's Kimi K3 (max) at 44-45. On output speed, Google's Gemini 3.5 Flash-Lite leads at 358 tokens per second and Gemini 3.8 Flash (high) at 298 t/s, with Meta's Muse Spark 1.3 (max) and DeepSeek's V4 Flash Vision (max) at 214 t/s.", "body_md": "# AI Coding Benchmarks & Rankings Leaderboard\n\nEmpirical evaluations measuring coding proficiency, reasoning benchmarks (GPQA, SWE-bench Verified, HLE), and real-time value-for-money ($/point) metrics.\n\n```\n                    Highlights\n                \n\n                    Updated Daily • Empirical Telemetry\n                \n\nIntelligence\n\nUpdated\n\n                            Artificial Analysis Intelligence Index • Higher is better\n                        \n\nGPT-6 Astra (max) Benchmarks, Price & Specs (Oct 2026): 55 (OpenAI)\n\n                                55                            \n\n                                O                            \n\n                                    GPT-6 Astra (m..                                \n\nClaude Fable 5.1 (max with fallback): 53 (Anthropic)\n\n                                53                            \n\n                                A                            \n\n                                    Claude Fable 5..                                \n\nClaude Opus 5 (max) Benchmarks, Price & Specs (Oct 2026): 53 (Anthropic)\n\n                                53                            \n\n                                A                            \n\n                                    Claude Opus 5 ..                                \n\nMuse Spark 1.3 (max): 48 (Meta)\n\n                                48                            \n\n                                M                            \n\n                                    Muse Spark 1.3..                                \n\nGPT-5.6 Sol (max) Benchmarks, Price & Specs (Oct 2026): 48 (OpenAI)\n\n                                48                            \n\n                                O                            \n\n                                    GPT-5.6 Sol (m..                                \n\nGLM-5.3 (max): 45 (Z AI)\n\n                                45                            \n\n                                Z                            \n\n                                    GLM-5.3 (max)                                \n\nQwen3.8 Max (0902): 45 (Alibaba)\n\n                                45                            \n\n                                Q                            \n\n                                    Qwen3.8 Max (0..                                \n\nStep 5 Preview: 44 (StepFun)\n\n                                44                            \n\n                                S                            \n\n                                    Step 5 Preview                                \n\nGrok 4.6 (high): 44 (SpaceXAI)\n\n                                44                            \n\n                                X                            \n\n                                    Grok 4.6 (high)                                \n\nKimi K3 (max): 44 (Kimi)\n\n                                44                            \n\n                                K                            \n\n                                    Kimi K3 (max)                                \n\nSpeed\n\n                            Output tokens per second • Higher is better\n                        \n\nGemini 3.5 Flash-Lite: 358 t/s (Google)\n\n                                358 t/s                            \n\n                                G                            \n\n                                    Gemini 3.5 Fla..                                \n\nGemini 3.8 Flash (high): 298 t/s (Google)\n\n                                298 t/s                            \n\n                                G                            \n\n                                    Gemini 3.8 Fla..                                \n\nMuse Spark 1.3 (max): 214 t/s (Meta)\n\n                                214 t/s                            \n\n                                M                            \n\n                                    Muse Spark 1.3..                                \n\nDeepSeek V4 Flash Vision (max): 214 t/s (DeepSeek)\n\n                                214 t/s                            \n\n                                D                            \n\n                                    DeepSeek V4 Fl..                                \n\nDeepSeek V4.1 Flash (max): 208 t/s (DeepSeek)\n\n                                208 t/s                            \n\n                                D                            \n\n                                    DeepSeek V4.1 ..                                \n\nMiniMax-M3: 186 t/s (MiniMax)\n\n                                186 t/s                            \n\n                                M                            \n\n                                    MiniMax-M3                                \n\ngpt-oss-120b (high): 181 t/s (OpenAI)\n\n                                181 t/s                            \n\n                                O                            \n\n                                    gpt-oss-120b (..                                \n\nNemotron 3 Ultra: 164 t/s (NVIDIA)\n\n                                164 t/s                            \n\n                                N                            \n\n                                    Nemotron 3 Ultra                                \n\nMistral Medium 3.5: 137 t/s (Mistral)\n\n                                137 t/s                            \n\n                                M                            \n\n                                    Mistral Medium..                                \n\nGPT-5.6 Luna (max): 133 t/s (OpenAI)\n\n                                133 t/s                            \n\n                                O                            \n\n                                    GPT-5.6 Luna (..                                \n\nCost per 1M Tokens\n\n                            Weighted average blended price (USD) • Lower is better\n                        \n\nMuse Glimmer (high): $0.12 (Meta)\n\n                                $0.12                            \n\n                                M                            \n\n                                    Muse Glimmer (..                                \n\nGemini 3.5 Flash-Lite: $0.18 (Google)\n\n                                $0.18                            \n\n                                G                            \n\n                                    Gemini 3.5 Fla..                                \n\ngpt-oss-120b (high): $0.18 (OpenAI)\n\n                                $0.18                            \n\n                                O                            \n\n                                    gpt-oss-120b (..                                \n\nGLM-5.3-Flash: $0.35 (Z AI)\n\n                                $0.35                            \n\n                                Z                            \n\n                                    GLM-5.3-Flash                                \n\nDeepSeek V4.1 Flash (max): $0.37 (DeepSeek)\n\n                                $0.37                            \n\n                                D                            \n\n                                    DeepSeek V4.1 ..                                \n\nQwen3.8-Flash-Next: $0.45 (Alibaba)\n\n                                $0.45                            \n\n                                Q                            \n\n                                    Qwen3.8-Flash-..                                \n\nDeepSeek V4 Flash Vision (max): $0.47 (DeepSeek)\n\n                                $0.47                            \n\n                                D                            \n\n                                    DeepSeek V4 Fl..                                \n\nGPT-5.6 Luna (max): $0.52 (OpenAI)\n\n                                $0.52                            \n\n                                O                            \n\n                                    GPT-5.6 Luna (..                                \n\nMistral Medium 3.5: $0.60 (Mistral)\n\n                                $0.60                            \n\n                                M                            \n\n                                    Mistral Medium..                                \n\nMiniMax-M3: $0.60 (MiniMax)\n\n                                $0.60                            \n\n                                M                            \n\n                                    MiniMax-M3\n```\n\n## Artificial Analysis Intelligence Index\n\n↗\nIntelligence Index v4.3 incorporates evaluations: SWE-bench Verified (40%), Terminal-Bench 2.0 (25%), Aider Polyglot (20%), LiveCodeBench (15%). Every figure is verified against primary source logs without prompt leakage.\n\n### Intelligence Index vs. Cost per 1M Tokens\n\nCode Intelligence Index vs. Weighted average cost (USD, Log Scale)\n\n### Frontier Language Model Intelligence, Over Time\n\nProgressive evolution across Anthropic, OpenAI, DeepSeek, Google, and Meta from 2023 to 2026.\n\n### Canonical Benchmark Suite\n\nBreakdown across isolated execution harnesses matching Artificial Analysis methodologies.\n\n### Value-for-Money Intelligence vs. Cost Matrix\n\n| RANK | MODEL & CREATOR |  |  |  |  |  |  |  | \n|---|---|---|---|---|---|---|---|---|\n| 01 | [Claude Fable 5.1 (max with fallback)](https://vibecoderjournal.com/rankings/claude-fable-5-1-max/) Anthropic • API • ▲ +1 |  | —% | —% | 1M | 68 t/s TTFT: 262210ms | $30.0000 | — | \n| 02 | [Muse Spark 1.3 (max)](https://vibecoderjournal.com/rankings/muse-spark-1-3-max/) Meta • API • ▲ +1 |  | —% | —% | 1M | 214 t/s TTFT: 24160ms | $2.2500 | — | \n| 03 | [GLM-5.3 (max)](https://vibecoderjournal.com/rankings/glm-5-3-max/) Z AI • API • ▲ +1 |  | —% | —% | 1M | 72 t/s TTFT: 2960ms | $1.5000 | — | \n| 04 | [Qwen3.8 Max (0902)](https://vibecoderjournal.com/rankings/qwen-3-8-max-0902/) Alibaba • API • ▲ +1 |  | —% | —% | 984K | 37 t/s TTFT: 2720ms | $3.0000 | — | \n| 05 | [Step 5 Preview](https://vibecoderjournal.com/rankings/step-5-preview/) StepFun • API • ▲ +1 |  | —% | —% | 1M | 100 t/s TTFT: 2960ms | $1.2000 | — | \n| 06 | [Grok 4.6 (high)](https://vibecoderjournal.com/rankings/grok-4-6-high/) SpaceXAI • API • ▲ +1 |  | —% | —% | 500K | 60 t/s TTFT: 44250ms | $4.0000 | — | \n| 07 | [Kimi K3 (max)](https://vibecoderjournal.com/rankings/kimi-k3-max/) Kimi • API • ▲ +1 |  | —% | —% | 1.05M | 38 t/s TTFT: 3990ms | $1.8000 | — | \n| 08 | [GPT-5.6 Terra (max)](https://vibecoderjournal.com/rankings/gpt-5-6-terra-max/) OpenAI • API • ▲ +1 |  | —% | —% | 1M | 84 t/s TTFT: 220860ms | $3.1500 | — | \n| 09 | [GLM-5.3-Flash](https://vibecoderjournal.com/rankings/glm-5-3-flash/) Z AI • API • ▲ +1 |  | —% | —% | 1M | 95 t/s TTFT: 2500ms | $0.3500 | — | \n| 10 | [Gemini 3.8 Flash (high)](https://vibecoderjournal.com/rankings/gemini-3-8-flash-high/) Google • API • ▲ +1 |  | —% | —% | 1M | 298 t/s TTFT: 15470ms | $1.3100 | — | \n| 11 | [DeepSeek V4.1 Flash (max)](https://vibecoderjournal.com/rankings/deepseek-v4-1-flash-max/) DeepSeek • API • ▲ +1 |  | —% | —% | 1M | 208 t/s TTFT: 1140ms | $0.3700 | — | \n| 12 | [Qwen3.8-Flash-Next](https://vibecoderjournal.com/rankings/qwen-3-8-flash-next/) Alibaba • API • ▲ +1 |  | —% | —% | 256K | 55 t/s TTFT: 2890ms | $0.4500 | — | \n| 13 | [Qwen3.8 2.4T A95B](https://vibecoderjournal.com/rankings/qwen-3-8-2-4t-a95b/) Alibaba • API • ▲ +1 |  | —% | —% | 984K | 38 t/s TTFT: 2690ms | $2.2500 | — | \n| 14 | [Claude Sonnet 5 (max)](https://vibecoderjournal.com/rankings/claude-sonnet-5-max/) Anthropic • API • ▲ +1 |  | —% | —% | 1M | 74 t/s TTFT: 133400ms | $8.0000 | — | \n| 15 | [GPT-5.6 Luna (max)](https://vibecoderjournal.com/rankings/gpt-5-6-luna-max/) OpenAI • API • ▲ +1 |  | —% | —% | 1M | 133 t/s TTFT: 113240ms | $0.5200 | — | \n| 16 | [DeepSeek V4 Flash Vision (max)](https://vibecoderjournal.com/rankings/deepseek-v4-flash-vision-max/) DeepSeek • API • ▲ +1 |  | —% | —% | 1M | 214 t/s TTFT: 1050ms | $0.4700 | — | \n| 17 | [Qwen3.8 27B (xhigh)](https://vibecoderjournal.com/rankings/qwen-3-8-27b-xhigh/) Alibaba • API • ▲ +1 |  | —% | —% | 256K | 43 t/s TTFT: 3840ms | $0.9000 | — | \n| 18 | [Nemotron 3 Ultra](https://vibecoderjournal.com/rankings/nemotron-3-ultra/) NVIDIA • API • ▲ +1 |  | —% | —% | 262K | 164 t/s TTFT: 2400ms | $0.7500 | — | \n| 19 | [gpt-oss-120b (high)](https://vibecoderjournal.com/rankings/gpt-oss-120b-high/) OpenAI • API • ▲ +1 |  | —% | —% | 131K | 181 t/s TTFT: 850ms | $0.1800 | — | \n| 20 | [Mistral Medium 3.5](https://vibecoderjournal.com/rankings/mistral-medium-3-5/) Mistral • API • ▼ -1 |  | —% | —% | 256K | 137 t/s TTFT: 2290ms | $0.6000 | — | \n| 21 | [Muse Glimmer (high)](https://vibecoderjournal.com/rankings/muse-glimmer-high/) Meta • API • ▼ -2 |  | —% | —% | 131K | 93 t/s TTFT: 990ms | $0.1200 | — | \n| 22 | [Gemini 3.5 Flash-Lite](https://vibecoderjournal.com/rankings/gemini-3-5-flash-lite/) Google • API • ▼ -3 |  | —% | —% | 1M | 358 t/s TTFT: 9780ms | $0.1750 | — | \n| 23 | [Inkling](https://vibecoderjournal.com/rankings/inkling/) Thinking Machines • API • ▼ -5 |  | —% | —% | 1M | 81 t/s TTFT: 2480ms | $0.7500 | — | \n| 24 | [Gemini 3.1 Pro Benchmarks, Price & Specs (Oct 2026)](https://vibecoderjournal.com/rankings/gemini-3-1-pro-preview/) Google • API • ▲ +1 |  | —% | —% | 1M | 119 t/s TTFT: 33160ms | $1.4000 | — | \n| 25 | [MiniMax-M3](https://vibecoderjournal.com/rankings/minimax-m3/) MiniMax • API • ▼ -7 |  | —% | —% | 1M | 186 t/s TTFT: 1000ms | $0.6000 | — | \n| 26 | [K2 Horizon 375B A23B](https://vibecoderjournal.com/rankings/k2-horizon-375b-a23b/) Institute of Foundation Models • API • ▼ -8 |  | —% | —% | 524K | 80 t/s TTFT: 3500ms | $1.2000 | — | \n| 27 | [DeepSeek V4 Pro Benchmarks, Price & Specs (Oct 2026)](https://vibecoderjournal.com/rankings/deepseek-v4-pro-0813-max/) DeepSeek • API • ▼ -11 |  | —% | —% | 1M | 88 t/s TTFT: 1700ms | $0.9600 | — | \n| 28 | [GPT-5.6 Sol (max) Benchmarks, Price & Specs (Oct 2026)](https://vibecoderjournal.com/rankings/gpt-5-6-sol-max/) OpenAI • API • ▼ -25 |  | —% | —% | 1M | 61 t/s TTFT: 129500ms | $6.1250 | — | \n| 29 | [Claude Opus 5 (max) Benchmarks, Price & Specs (Oct 2026)](https://vibecoderjournal.com/rankings/claude-opus-5-max/) Anthropic • API • ▲ +1 |  | —% | —% | 1M | 51 t/s TTFT: 57030ms | $20.0000 | — | \n| 30 | [GPT-6 Astra (max) Benchmarks, Price & Specs (Oct 2026)](https://vibecoderjournal.com/rankings/gpt-6-astra-max/) OpenAI • API • ▼ -29 |  | —% | —% | 1M | 57 t/s TTFT: 283660ms | $12.0000 | — | \n\n**53.1%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**68 t/s (262210ms)**\n\n**$30.0000**\n\n**—**\n\n**48.0%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**214 t/s (24160ms)**\n\n**$2.2500**\n\n**—**\n\n[GLM-5.3 (max)](https://vibecoderjournal.com/rankings/glm-5-3-max/)\n\n**45.1%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**72 t/s (2960ms)**\n\n**$1.5000**\n\n**—**\n\n**45.0%**\n\n**—%**\n\n**—%**\n\n**984K**\n\n**37 t/s (2720ms)**\n\n**$3.0000**\n\n**—**\n\n[Step 5 Preview](https://vibecoderjournal.com/rankings/step-5-preview/)\n\n**44.1%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**100 t/s (2960ms)**\n\n**$1.2000**\n\n**—**\n\n[Grok 4.6 (high)](https://vibecoderjournal.com/rankings/grok-4-6-high/)\n\n**44.0%**\n\n**—%**\n\n**—%**\n\n**500K**\n\n**60 t/s (44250ms)**\n\n**$4.0000**\n\n**—**\n\n[Kimi K3 (max)](https://vibecoderjournal.com/rankings/kimi-k3-max/)\n\n**44.0%**\n\n**—%**\n\n**—%**\n\n**1.05M**\n\n**38 t/s (3990ms)**\n\n**$1.8000**\n\n**—**\n\n**42.1%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**84 t/s (220860ms)**\n\n**$3.1500**\n\n**—**\n\n[GLM-5.3-Flash](https://vibecoderjournal.com/rankings/glm-5-3-flash/)\n\n**42.1%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**95 t/s (2500ms)**\n\n**$0.3500**\n\n**—**\n\n**41.0%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**298 t/s (15470ms)**\n\n**$1.3100**\n\n**—**\n\n**40.0%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**208 t/s (1140ms)**\n\n**$0.3700**\n\n**—**\n\n**40.0%**\n\n**—%**\n\n**—%**\n\n**256K**\n\n**55 t/s (2890ms)**\n\n**$0.4500**\n\n**—**\n\n**40.0%**\n\n**—%**\n\n**—%**\n\n**984K**\n\n**38 t/s (2690ms)**\n\n**$2.2500**\n\n**—**\n\n**38.1%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**74 t/s (133400ms)**\n\n**$8.0000**\n\n**—**\n\n**38.0%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**133 t/s (113240ms)**\n\n**$0.5200**\n\n**—**\n\n**35.0%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**214 t/s (1050ms)**\n\n**$0.4700**\n\n**—**\n\n**34.0%**\n\n**—%**\n\n**—%**\n\n**256K**\n\n**43 t/s (3840ms)**\n\n**$0.9000**\n\n**—**\n\n[Nemotron 3 Ultra](https://vibecoderjournal.com/rankings/nemotron-3-ultra/)\n\n**23.0%**\n\n**—%**\n\n**—%**\n\n**262K**\n\n**164 t/s (2400ms)**\n\n**$0.7500**\n\n**—**\n\n**0.0%**\n\n**—%**\n\n**—%**\n\n**131K**\n\n**181 t/s (850ms)**\n\n**$0.1800**\n\n**—**\n\n**0.0%**\n\n**—%**\n\n**—%**\n\n**256K**\n\n**137 t/s (2290ms)**\n\n**$0.6000**\n\n**—**\n\n**0.0%**\n\n**—%**\n\n**—%**\n\n**131K**\n\n**93 t/s (990ms)**\n\n**$0.1200**\n\n**—**\n\n**0.0%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**358 t/s (9780ms)**\n\n**$0.1750**\n\n**—**\n\n[Inkling](https://vibecoderjournal.com/rankings/inkling/)\n\n**0.0%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**81 t/s (2480ms)**\n\n**$0.7500**\n\n**—**\n\n**0.0%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**119 t/s (33160ms)**\n\n**$1.4000**\n\n**—**\n\n[MiniMax-M3](https://vibecoderjournal.com/rankings/minimax-m3/)\n\n**0.0%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**186 t/s (1000ms)**\n\n**$0.6000**\n\n**—**\n\n**0.0%**\n\n**—%**\n\n**—%**\n\n**524K**\n\n**80 t/s (3500ms)**\n\n**$1.2000**\n\n**—**\n\n**0.0%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**88 t/s (1700ms)**\n\n**$0.9600**\n\n**—**\n\n**0.0%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**61 t/s (129500ms)**\n\n**$6.1250**\n\n**—**\n\n**0.0%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**51 t/s (57030ms)**\n\n**$20.0000**\n\n**—**\n\n**0.0%**\n\n**—%**\n\n**—%**\n\n**1M**\n\n**57 t/s (283660ms)**\n\n**$12.0000**\n\n**—**", "url": "https://wpnews.pro/news/show-hn-independent-llm-benchmark-index-tracking-22-frontier-models", "canonical_source": "https://vibecoderjournal.com/rankings/", "published_at": "2026-10-18 12:36:53+00:00", "updated_at": "2026-10-18 12:47:46.960937+00:00", "lang": "en", "topics": ["large-language-models", "ai-research", "artificial-intelligence", "generative-ai"], "entities": ["OpenAI", "Anthropic", "Meta", "Google", "DeepSeek", "Alibaba", "Z AI", "StepFun"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/show-hn-independent-llm-benchmark-index-tracking-22-frontier-models", "markdown": "https://wpnews.pro/news/show-hn-independent-llm-benchmark-index-tracking-22-frontier-models.md", "text": "https://wpnews.pro/news/show-hn-independent-llm-benchmark-index-tracking-22-frontier-models.txt", "jsonld": "https://wpnews.pro/news/show-hn-independent-llm-benchmark-index-tracking-22-frontier-models.jsonld"}}