Step 5 Preview vs Trinity Large Thinking

Mana yang lebih baik untuk uang: Step 5 Preview oleh StepFun atau Trinity Large Thinking oleh Arcee AI? Kecerdasan, total biaya benchmark, dan waktu end-to-end dari Artificial Analysis & sumber lain, cuplikan 2026-09-22.

Kompromi
StepFunStep 5 PreviewDi luar batas Pareto
Skor kecerdasan (Intelligence Index)
43.70
Biaya benchmark (USD)
$925
Arcee AITrinity Large ThinkingDi luar batas Pareto
Skor kecerdasan (Intelligence Index)
10.80
Biaya benchmark (USD)
$518
Kesimpulan

Tergantung anggaran Anda

Step 5 Preview mendapat 32.9 poin lebih tinggi (43.7 vs 10.8), tetapi biaya menjalankan benchmark-nya 1.8× lebih besar ($925 vs $518; keduanya kelas premium). Pilih Step 5 Preview jika kecerdasan tambahan sepadan dengan biaya lebihnya; jika tidak, Trinity Large Thinking.

MetrikStep 5 PreviewTrinity Large ThinkingSelisih
Kecerdasan (lebih tinggi lebih baik)43.710.8−75.3% Trinity Large Thinking 75.3% kurang cerdas
Total biaya AA (lebih rendah lebih baik)$924.68$518.07−44.0% Trinity Large Thinking 44.0% lebih murah
Waktu ujung ke ujung (lebih rendah lebih baik)33.68s9.39s−72.1% Trinity Large Thinking 72.1% lebih cepat
Harga per 1 juta token (masuk/keluar)$1.00 / $2.70$0.25 / $0.90
Jenis inputTeks, GambarTeks

Trinity Large Thinking dibanding Step 5 Preview: kecerdasan 75.3% lebih rendah, biaya menjalankan benchmark AA 44.0% lebih rendah, dan waktu total 72.1% lebih cepat.

Buka pertarungan ini di pemilih interaktifStep 5 Preview: model apa yang mengunggulinyaTrinity Large Thinking: model apa yang mengunggulinya
Pareto frontier hari ini

Dibandingkan dengan batas Pareto

0102030405060$20$50$100$200$500$1k$2k$5k$10k$20kBiaya untuk menjalankan Artificial Analysis Intelligence Index (USD, skala log)Indeks kecerdasanCeleris-1 (Celeris): kecerdasan 6.30, $60.0Claude 4.5 Haiku (Reasoning) (Anthropic): kecerdasan 16.90, $524Claude 4.5 Sonnet (Reasoning) (Anthropic): kecerdasan 20.70, $1501Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) (Anthropic): kecerdasan 49.60, $11161Claude Fable 5.1 (Adaptive Reasoning, High Effort, Default Fallback) (Anthropic): kecerdasan 51.20, $5242Claude Fable 5.1 (Adaptive Reasoning, Low Effort, Default Fallback) (Anthropic): kecerdasan 46.80, $3158Claude Fable 5.1 (Adaptive Reasoning, Medium Effort, Default Fallback) (Anthropic): kecerdasan 48.90, $3983Claude Opus 4.8 (Adaptive Reasoning, Max Effort) (Anthropic): kecerdasan 41.80, $6874Claude Opus 5 (Adaptive Reasoning, High Effort) (Anthropic): kecerdasan 48.10, $4332Claude Opus 5 (Adaptive Reasoning, Low Effort) (Anthropic): kecerdasan 39.40, $1561Claude Opus 5 (Adaptive Reasoning, Max Effort) (Anthropic): kecerdasan 50.80, $7275Claude Opus 5 (Adaptive Reasoning, Medium Effort) (Anthropic): kecerdasan 44.80, $2732Claude Opus 5 (Adaptive Reasoning, Xhigh Effort) (Anthropic): kecerdasan 49.70, $5868Claude Sonnet 4.6 (Adaptive Reasoning, Max Effort) (Anthropic): kecerdasan 30.10, $4925Claude Sonnet 5 (Adaptive Reasoning, High Effort) (Anthropic): kecerdasan 31.70, $2086Claude Sonnet 5 (Adaptive Reasoning, Low Effort) (Anthropic): kecerdasan 24.30, $653Claude Sonnet 5 (Adaptive Reasoning, Max Effort) (Anthropic): kecerdasan 38.20, $6998Claude Sonnet 5 (Adaptive Reasoning, Medium Effort) (Anthropic): kecerdasan 28.10, $1218Claude Sonnet 5 (Adaptive Reasoning, Xhigh Effort) (Anthropic): kecerdasan 34.40, $3255DeepSeek R1 (Jan '25) (DeepSeek): kecerdasan 11.40, $461DeepSeek V3 (Dec '24) (DeepSeek): kecerdasan 8.50, $25.4DeepSeek V3 0324 (DeepSeek): kecerdasan 9.70, $87.2DeepSeek V4 Flash 0420 (Reasoning, Max Effort) (DeepSeek): kecerdasan 24.20, $191DeepSeek V4 Flash 0731 (Reasoning, Max Effort) (DeepSeek): kecerdasan 34.30, $474DeepSeek V4 Flash Vision (Reasoning, Max Effort) (DeepSeek): kecerdasan 34.80, $445DeepSeek V4 Pro 0424 (Reasoning, Max Effort) (DeepSeek): kecerdasan 30.40, $277DeepSeek V4 Pro 0813 (Reasoning, Max Effort) (DeepSeek): kecerdasan 36.00, $1122DeepSeek V4.1 Flash (Reasoning, Max Effort) (DeepSeek): kecerdasan 39.50, $477Gemini 2.5 Pro (Google): kecerdasan 16.10, $707Gemini 3.1 Flash-Lite (Google): kecerdasan 15.60, $119Gemini 3.1 Pro Preview (Google): kecerdasan 29.70, $1310Gemini 3.5 Flash (high) (Google): kecerdasan 32.60, $2172Gemini 3.5 Flash-Lite (Google): kecerdasan 22.20, $267Gemini 3.6 Flash (high) (Google): kecerdasan 34.00, $1037Gemini 3.7 Flash (high) (Google): kecerdasan 39.10, $1084Gemini 3.8 Flash (high) (Google): kecerdasan 40.90, $1623Gemini 3.8 Flash (medium) (Google): kecerdasan 39.80, $1100GLM 5.3 Flash (Z AI): kecerdasan 41.80, $280GLM-5.1 (Reasoning) (Z AI): kecerdasan 26.10, $1418GLM-5.2 (max) (Z AI): kecerdasan 33.70, $1559GLM-5.3 (max) (Z AI): kecerdasan 44.80, $2503GPT-5 mini (high) (OpenAI): kecerdasan 16.80, $189GPT-5.4 mini (xhigh) (OpenAI): kecerdasan 24.10, $1438GPT-5.4 nano (xhigh) (OpenAI): kecerdasan 20.70, $331GPT-5.5 (high) (OpenAI): kecerdasan 37.00, $3411GPT-5.5 (medium) (OpenAI): kecerdasan 33.80, $2095GPT-5.5 (xhigh) (OpenAI): kecerdasan 38.40, $5294GPT-5.5 Instant (June 2026) (OpenAI): kecerdasan 26.00, $1184GPT-5.6 Luna (max) (OpenAI): kecerdasan 37.30, $320GPT-5.6 Sol (high) (OpenAI): kecerdasan 42.30, $1487GPT-5.6 Sol (low) (OpenAI): kecerdasan 33.50, $637GPT-5.6 Sol (max) (OpenAI): kecerdasan 47.00, $3465GPT-5.6 Sol (medium) (OpenAI): kecerdasan 39.20, $997GPT-5.6 Sol (xhigh) (OpenAI): kecerdasan 44.00, $2082GPT-5.6 Terra (high) (OpenAI): kecerdasan 34.20, $768GPT-5.6 Terra (low) (OpenAI): kecerdasan 27.50, $350GPT-5.6 Terra (max) (OpenAI): kecerdasan 42.10, $2501GPT-5.6 Terra (medium) (OpenAI): kecerdasan 30.10, $447GPT-5.6 Terra (Non-reasoning) (OpenAI): kecerdasan 20.80, $256GPT-5.6 Terra (xhigh) (OpenAI): kecerdasan 38.00, $1187GPT-6 Astra (low) (OpenAI): kecerdasan 45.80, $1537gpt-oss-120b (high) (OpenAI): kecerdasan 11.60, $122gpt-oss-20b (high) (OpenAI): kecerdasan 9.00, $29.4Granite 4.2 8B (IBM): kecerdasan 11.10, $53.1Grok 4.3 (high) (SpaceXAI): kecerdasan 24.90, $332Grok 4.3 (Non-reasoning) (SpaceXAI): kecerdasan 14.00, $203Grok 4.5 (high) (SpaceXAI): kecerdasan 38.80, $1404Grok 4.6 (high) (SpaceXAI): kecerdasan 44.30, $2352Grok 4.6 (low) (SpaceXAI): kecerdasan 35.10, $761Grok 4.6 (medium) (SpaceXAI): kecerdasan 42.80, $1937Grok 4.6 (xhigh) (SpaceXAI): kecerdasan 44.20, $2830Grok 4.7 (high) (SpaceXAI): kecerdasan 46.30, $3881Grok 4.7 (xhigh) (SpaceXAI): kecerdasan 46.40, $4967Hy3 (Tencent): kecerdasan 25.30, $156Inkling (xhigh) (Thinking Machines): kecerdasan 25.00, $1026Kimi K2.6 (Kimi): kecerdasan 27.00, $1528Kimi K2.7 Code (Kimi): kecerdasan 25.80, $993Kimi K3 (max) (Kimi): kecerdasan 43.60, $3658LongCat 2.0 (LongCat): kecerdasan 19.10, $221MiMo-V2.5-Pro (Xiaomi): kecerdasan 26.00, $138MiniMax-M2.7 (MiniMax): kecerdasan 22.80, $236MiniMax-M3 (MiniMax): kecerdasan 29.20, $538Ministral 3 14B (Mistral): kecerdasan 6.00, $250Ministral 3 3B (Mistral): kecerdasan 4.80, $61.3Ministral 3 8B (Mistral): kecerdasan 5.50, $169Mistral Large 3 (Mistral): kecerdasan 9.30, $130Mistral Medium 3.1 (Mistral): kecerdasan 9.20, $406Mistral Medium 3.5 (Mistral): kecerdasan 14.20, $1160Mistral Small 3.1 (Mistral): kecerdasan 7.10, $35.8Mistral Small 4 (Reasoning) (Mistral): kecerdasan 11.30, $78.7Muse Glimmer (high) (Meta): kecerdasan 17.50, $152Muse Spark 1.1 (xhigh) (Meta): kecerdasan 33.70, $1737Muse Spark 1.2 (xhigh) (Meta): kecerdasan 39.60, $1385Muse Spark 1.3 (xhigh) (Meta): kecerdasan 45.10, $1655Nemotron 3 Super 120B A12B (Reasoning) (NVIDIA): kecerdasan 12.80, $1319Nemotron 3 Ultra 550B A55B (Reasoning) (NVIDIA): kecerdasan 22.90, $778Nemotron 3.5 Lightning (NVIDIA): kecerdasan 12.90, $105NVIDIA Nemotron 3 Nano 30B A3B (Reasoning) (NVIDIA): kecerdasan 8.90, $63.7Quasar 438B (max, based on GLM-5.2) (Multiverse Computing): kecerdasan 26.70, $2504Qwen3 235B A22B 2507 (Reasoning) (Alibaba): kecerdasan 12.70, $235Qwen3 30B A3B 2507 (Reasoning) (Alibaba): kecerdasan 9.80, $188Qwen3 Coder Next (Alibaba): kecerdasan 9.20, $476Qwen3.5 122B A10B (Reasoning) (Alibaba): kecerdasan 15.60, $661Qwen3.5 397B A17B (Reasoning) (Alibaba): kecerdasan 18.40, $941Qwen3.6 27B (Reasoning) (Alibaba): kecerdasan 21.40, $1145Qwen3.6 35B A3B (Reasoning) (Alibaba): kecerdasan 18.20, $896Qwen3.7 Max (Alibaba): kecerdasan 29.50, $2277Qwen3.7 Plus (Alibaba): kecerdasan 25.20, $568Qwen3.8 2.4T A95B (Alibaba): kecerdasan 39.90, $2633Qwen3.8 27B (low) (Alibaba): kecerdasan 26.20, $876Qwen3.8 27B (medium) (Alibaba): kecerdasan 27.60, $977Qwen3.8 27B (Non-reasoning) (Alibaba): kecerdasan 20.20, $1570Qwen3.8 27B (xhigh) (Alibaba): kecerdasan 33.70, $1170Qwen3.8 Max (Alibaba): kecerdasan 40.20, $3172Qwen3.8 Max (0902) (Alibaba): kecerdasan 45.40, $4935Qwen3.8-Flash-Next (Alibaba): kecerdasan 39.80, $363Ring-2.6-1T (InclusionAI): kecerdasan 16.60, $576Solar Pro 3 (Upstage): kecerdasan 7.80, $159Granite 4.2 3B (IBM): kecerdasan 9.10, $14.9GPT-5.6 Luna (Non-reasoning) (OpenAI): kecerdasan 15.50, $23.0GPT-5.6 Luna (low) (OpenAI): kecerdasan 21.00, $27.0GPT-5.6 Luna (medium) (OpenAI): kecerdasan 25.00, $44.3GPT-5.6 Luna (high) (OpenAI): kecerdasan 32.10, $108GPT-5.6 Luna (xhigh) (OpenAI): kecerdasan 34.60, $179MiMo-V2.6-Pro (Xiaomi): kecerdasan 46.30, $207Muse Spark 1.3 (max) (Meta): kecerdasan 48.10, $2000GPT-6 Astra (medium) (OpenAI): kecerdasan 49.60, $2434GPT-6 Astra (high) (OpenAI): kecerdasan 50.90, $2925GPT-6 Astra (xhigh) (OpenAI): kecerdasan 52.40, $3803GPT-6 Astra (max) (OpenAI): kecerdasan 52.70, $5324Claude Fable 5.1 (Adaptive Reasoning, Xhigh Effort, Default Fallback) (Anthropic): kecerdasan 53.20, $9063Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback) (Anthropic): kecerdasan 53.40, $13129Trinity Large Thinking (Arcee AI): kecerdasan 10.80, $518Step 5 Preview (StepFun): kecerdasan 43.70, $925Step 5 PreviewTrinity Large ThinkingGranite 4.2 3BClaude Fable 5.1 (max)
Step 5 PreviewTrinity Large ThinkingPareto frontier hari iniBatas Pareto model dengan bobot yang dapat diunduh (open-weights frontier)Semua model lainnya
Perbandingan langsung

Bandingkan sesuatu yang lain

Agen: Accept: text/markdown or ?format=md mengembalikan halaman ini sebagai Markdown. JSON: /api/v1/compare/step-5/trinity-large-thinking · Markdown: /api/v1/compare/step-5/trinity-large-thinking?format=md · llms.txt · keterampilan agen