Sari la conținut
aiqbench

Română · beta. Traducerea a fost făcută automat și este încă în curs de revizuire. Citește pagina în engleză

Frontiera capabilităților în timp

Fiecare benchmark Epoch, reprezentat la data de lansare a fiecărui model evaluat pe el. Pe măsură ce apar modele noi, plafonul a ceea ce s-a realizat continuă să urce — o tendință reală, distinctă de (și adesea confundată cu) îmbunătățirea în timp a modelelor individuale.

BenchmarkGPQA DiamondOTIS Mock AIMEFrontierMathSimpleQA VerifiedSWE-bench VerifiedARC-AGI-2Terminal-BenchHumanity's Last ExamSimpleBench

GPQA Diamond — modelele la apariție

Un punct per model, la data lui de lansare. Linia urmărește cel mai bun scor cunoscut în fiecare moment.

0%25%50%75%100%20232024202520262027Data de lansare a modeluluiScor (%)gpt-4-0314 Lansat: mar. 2023 Scor: 35.7% CI 95%: [33.4%, 38.1%] Sursa datei: curatedgpt-4-0613 Lansat: iun. 2023 Scor: 30.7% CI 95%: [28.3%, 33.0%] Sursa datei: curatedYi-34B-Chat Lansat: nov. 2023 Scor: 14.7% CI 95%: [13.6%, 15.9%] Sursa datei: curatedgpt-3.5-turbo-1106 Lansat: nov. 2023 Scor: 28.0% CI 95%: [26.2%, 29.9%] Sursa datei: curatedgpt-4-1106-preview Lansat: nov. 2023 Scor: 42.4% CI 95%: [40.0%, 44.8%] Sursa datei: curateddeepseek-llm-67b-chat Lansat: nov. 2023 Scor: 24.6% CI 95%: [23.3%, 26.0%] Sursa datei: curatedMixtral-8x7B-Instruct-v0.1 Lansat: dec. 2023 Scor: 30.6% CI 95%: [28.7%, 32.5%] Sursa datei: curatedgpt-4-0125-preview Lansat: ian. 2024 Scor: 42.3% CI 95%: [39.6%, 44.9%] Sursa datei: curatedgpt-3.5-turbo-0125 Lansat: ian. 2024 Scor: 27.2% CI 95%: [25.2%, 29.1%] Sursa datei: curatedmistral-large-2402 Lansat: feb. 2024 Scor: 38.8% CI 95%: [36.3%, 41.2%] Sursa datei: name-suffixqwen1.5-72b-chat Lansat: feb. 2024 Scor: 28.8% CI 95%: [26.8%, 30.8%] Sursa datei: curatedclaude-3-opus-20240229 Lansat: mar. 2024 Scor: 47.2% CI 95%: [44.5%, 49.8%] Sursa datei: curatedclaude-3-sonnet-20240229 Lansat: mar. 2024 Scor: 40.6% CI 95%: [38.2%, 43.0%] Sursa datei: curatedclaude-3-haiku-20240307 Lansat: mar. 2024 Scor: 36.3% CI 95%: [33.9%, 38.7%] Sursa datei: curatedqwen1.5-32b-chat Lansat: apr. 2024 Scor: 30.7% CI 95%: [28.7%, 32.8%] Sursa datei: curatedgemini-1.5-pro-001 Lansat: apr. 2024 Scor: 45.9% CI 95%: [43.2%, 48.5%] Sursa datei: curatedgpt-4-turbo-2024-04-09 Lansat: apr. 2024 Scor: 46.6% CI 95%: [43.9%, 49.2%] Sursa datei: curatedWizardLM-2-8x22B Lansat: apr. 2024 Scor: 43.4% CI 95%: [41.1%, 45.8%] Sursa datei: curatedYi-1.5-34B-Chat Lansat: mai 2024 Scor: 32.0% CI 95%: [30.0%, 34.0%] Sursa datei: curatedgpt-4o-2024-08-06 Lansat: mai 2024 Scor: 49.2% CI 95%: [46.6%, 51.8%] Sursa datei: curatedgemini-1.5-flash-001 Lansat: mai 2024 Scor: 40.4% CI 95%: [38.0%, 42.8%] Sursa datei: curatedqwen2-72b-instruct Lansat: iun. 2024 Scor: 40.8% CI 95%: [38.2%, 43.4%] Sursa datei: curatedclaude-3-5-sonnet-20241022 Lansat: iun. 2024 Scor: 55.3% CI 95%: [52.5%, 58.1%] Sursa datei: curatedgemma-2-9b-it Lansat: iun. 2024 Scor: 27.5% CI 95%: [25.2%, 29.8%] Sursa datei: curatedgemma-2-27b-it Lansat: iun. 2024 Scor: 36.5% CI 95%: [34.1%, 38.9%] Sursa datei: curatedmistral-large-2407 Lansat: iul. 2024 Scor: 49.0% CI 95%: [46.4%, 51.6%] Sursa datei: name-suffixgpt-4o-mini-2024-07-18 Lansat: iul. 2024 Scor: 37.7% CI 95%: [35.3%, 40.1%] Sursa datei: curatedLlama-3.1-8B-Instruct Lansat: iul. 2024 Scor: 27.0% CI 95%: [25.1%, 28.8%] Sursa datei: curatedLlama-3.1-70B-Instruct Lansat: iul. 2024 Scor: 44.2% CI 95%: [41.7%, 46.6%] Sursa datei: curatedo1-preview-2024-09-12 Lansat: sept. 2024 Scor: 50.3% CI 95%: [47.5%, 53.1%] Sursa datei: curatedo1-mini-2024-09-12 Lansat: sept. 2024 Scor: 62.4% CI 95%: [59.6%, 65.1%] Sursa datei: curatedqwen2.5-72b-instruct Lansat: sept. 2024 Scor: 49.1% CI 95%: [46.5%, 51.8%] Sursa datei: curatedgemini-1.5-flash-002 Lansat: sept. 2024 Scor: 47.3% CI 95%: [44.5%, 50.1%] Sursa datei: curatedgemini-1.5-pro-002 Lansat: sept. 2024 Scor: 57.2% CI 95%: [54.4%, 60.0%] Sursa datei: curatedLlama-3.2-1B-Instruct Lansat: sept. 2024 Scor: 23.9% CI 95%: [22.3%, 25.6%] Sursa datei: curatedministral-8b-2410 Lansat: oct. 2024 Scor: 27.1% CI 95%: [25.1%, 29.2%] Sursa datei: name-suffixgemini-1.5-flash-8b-001 Lansat: oct. 2024 Scor: 33.0% CI 95%: [30.5%, 35.4%] Sursa datei: curatedmistral-large-2411 Lansat: nov. 2024 Scor: 51.3% CI 95%: [48.6%, 54.0%] Sursa datei: name-suffixclaude-3-5-haiku-20241022 Lansat: nov. 2024 Scor: 38.1% CI 95%: [35.5%, 40.8%] Sursa datei: curatedo1-2024-12-17 Lansat: dec. 2024 Scor: 76.8% CI 95%: [73.8%, 79.8%] Sursa datei: curatedLlama-3.3-70B-Instruct Lansat: dec. 2024 Scor: 47.4% CI 95%: [44.6%, 50.3%] Sursa datei: curatedphi-4 Lansat: dec. 2024 Scor: 56.1% CI 95%: [53.5%, 58.7%] Sursa datei: curateddeepseek-chat Lansat: dec. 2024 Scor: 71.2% CI 95%: [68.0%, 74.4%] Sursa datei: curatedDeepSeek-V3 Lansat: dec. 2024 Scor: 56.5% CI 95%: [53.8%, 59.3%] Sursa datei: curatedDeepSeek-R1-Distill-Llama-70B Lansat: ian. 2025 Scor: 55.7% CI 95%: [52.8%, 58.7%] Sursa datei: curatedDeepSeek-R1 Lansat: ian. 2025 Scor: 69.2% CI 95%: [66.2%, 72.3%] Sursa datei: curatedo3-mini-2025-01-31 Lansat: ian. 2025 Scor: 77.0% CI 95%: [74.4%, 79.6%] Sursa datei: curatedgemini-2.0-flash-001 Lansat: feb. 2025 Scor: 64.1% CI 95%: [61.4%, 66.9%] Sursa datei: curatedgrok-3-mini-beta Lansat: feb. 2025 Scor: 76.3% CI 95%: [73.2%, 79.3%] Sursa datei: curatedclaude-3-7-sonnet-20250219 Lansat: feb. 2025 Scor: 78.5% CI 95%: [75.9%, 81.2%] Sursa datei: curatedgpt-4.5-preview-2025-02-27 Lansat: feb. 2025 Scor: 68.7% CI 95%: [65.4%, 72.0%] Sursa datei: curatedQwQ-32B Lansat: mar. 2025 Scor: 65.3% CI 95%: [62.5%, 68.2%] Sursa datei: curatedgemma-3-4b-it Lansat: mar. 2025 Scor: 23.2% CI 95%: [21.4%, 25.1%] Sursa datei: curatedgemma-3-27b-it Lansat: mar. 2025 Scor: 47.7% CI 95%: [45.0%, 50.4%] Sursa datei: curatedgemma-3-12b-it Lansat: mar. 2025 Scor: 39.5% CI 95%: [36.9%, 42.0%] Sursa datei: curatedDeepSeek-V3-0324 Lansat: mar. 2025 Scor: 67.6% CI 95%: [64.9%, 70.3%] Sursa datei: curatedLlama-4-Scout-17B-16E-Instruct Lansat: apr. 2025 Scor: 51.8% CI 95%: [48.7%, 55.0%] Sursa datei: curatedgpt-4.1-mini-2025-04-14 Lansat: apr. 2025 Scor: 65.8% CI 95%: [63.1%, 68.6%] Sursa datei: curatedgpt-4.1-nano-2025-04-14 Lansat: apr. 2025 Scor: 48.9% CI 95%: [46.4%, 51.4%] Sursa datei: curatedgpt-4.1-2025-04-14 Lansat: apr. 2025 Scor: 66.9% CI 95%: [64.1%, 69.7%] Sursa datei: curatedo3-2025-04-16 Lansat: apr. 2025 Scor: 81.8% CI 95%: [79.7%, 83.9%] Sursa datei: curatedo4-mini-2025-04-16 Lansat: apr. 2025 Scor: 79.6% CI 95%: [77.2%, 82.0%] Sursa datei: curatedqwen-plus-2025-01-25 Lansat: apr. 2025 Scor: 48.1% CI 95%: [45.4%, 50.8%] Sursa datei: name-suffixqwen3-235b-a22b Lansat: apr. 2025 Scor: 70.7% CI 95%: [68.0%, 73.4%] Sursa datei: curatedqwen3-8b Lansat: apr. 2025 Scor: 56.8% CI 95%: [53.8%, 59.7%] Sursa datei: curatedQwen3-30B-A3B Lansat: apr. 2025 Scor: 61.7% CI 95%: [58.7%, 64.7%] Sursa datei: curatedqwen3-14b Lansat: apr. 2025 Scor: 63.8% CI 95%: [60.8%, 66.7%] Sursa datei: curatedQwen3-32B Lansat: apr. 2025 Scor: 65.7% CI 95%: [62.9%, 68.5%] Sursa datei: curatedmistral-medium-2505 Lansat: mai 2025 Scor: 59.5% CI 95%: [56.7%, 62.4%] Sursa datei: name-suffixgemini-2.5-pro-preview-05-06 Lansat: mai 2025 Scor: 66.7% CI 95%: [63.3%, 70.0%] Sursa datei: curatedclaude-sonnet-4-20250514 Lansat: mai 2025 Scor: 78.3% CI 95%: [75.3%, 81.2%] Sursa datei: curatedclaude-opus-4-20250514 Lansat: mai 2025 Scor: 76.3% CI 95%: [73.2%, 79.3%] Sursa datei: curatedDeepSeek-R1-0528 Lansat: mai 2025 Scor: 76.3% CI 95%: [73.9%, 78.7%] Sursa datei: curatedgemini-2.5-pro-preview-06-05 Lansat: iun. 2025 Scor: 84.8% CI 95%: [82.3%, 87.4%] Sursa datei: curatedgemini-2.5-pro Lansat: iun. 2025 Scor: 85.3% CI 95%: [83.2%, 87.4%] Sursa datei: curatedqwen3-30b-a3b-thinking-2507 Lansat: iul. 2025 Scor: 70.1% CI 95%: [67.2%, 72.9%] Sursa datei: name-suffixqwen3-30b-a3b-instruct-2507 Lansat: iul. 2025 Scor: 55.6% CI 95%: [52.7%, 58.5%] Sursa datei: name-suffixQwen3-235B-A22B-Thinking-2507 Lansat: iul. 2025 Scor: 80.1% CI 95%: [77.5%, 82.7%] Sursa datei: name-suffixgrok-4-0709 Lansat: iul. 2025 Scor: 87.0% CI 95%: [85.0%, 89.0%] Sursa datei: curatedclaude-opus-4-1-20250805 Lansat: aug. 2025 Scor: 77.3% CI 95%: [74.3%, 80.3%] Sursa datei: curatedgpt-oss-20b Lansat: aug. 2025 Scor: 60.8% CI 95%: [58.3%, 63.3%] Sursa datei: curatedgpt-5-2025-08-07 Lansat: aug. 2025 Scor: 86.2% CI 95%: [84.1%, 88.3%] Sursa datei: curatedgpt-5-mini-2025-08-07 Lansat: aug. 2025 Scor: 75.0% CI 95%: [72.7%, 77.3%] Sursa datei: curatedgpt-5-nano-2025-08-07 Lansat: aug. 2025 Scor: 69.4% CI 95%: [66.7%, 72.2%] Sursa datei: curatedqwen3-max-2025-09-23 Lansat: sept. 2025 Scor: 72.6% CI 95%: [69.9%, 75.3%] Sursa datei: curatedclaude-sonnet-4-5-20250929 Lansat: sept. 2025 Scor: 82.3% CI 95%: [79.6%, 85.0%] Sursa datei: curatedclaude-haiku-4-5-20251001 Lansat: oct. 2025 Scor: 71.2% CI 95%: [68.0%, 74.4%] Sursa datei: curatedclaude-opus-4-5-20251101 Lansat: nov. 2025 Scor: 86.0% CI 95%: [84.0%, 88.1%] Sursa datei: name-suffixkimi-k2-thinking-turbo Lansat: nov. 2025 Scor: 84.2% CI 95%: [82.1%, 86.3%] Sursa datei: curatedgpt-5.1-2025-11-13 Lansat: nov. 2025 Scor: 87.6% CI 95%: [85.8%, 89.5%] Sursa datei: name-suffixgpt-5.1gpt-5.2-2025-12-11 Lansat: dec. 2025 Scor: 91.4% CI 95%: [89.6%, 93.2%] Sursa datei: name-suffixgpt-5.2glm-4.7 Lansat: dec. 2025 Scor: 83.3% CI 95%: [81.0%, 85.7%] Sursa datei: huggingface-createdglm-4.7-flash Lansat: ian. 2026 Scor: 60.5% CI 95%: [57.6%, 63.5%] Sursa datei: huggingface-createdglm-5 Lansat: feb. 2026 Scor: 87.8% CI 95%: [85.5%, 90.2%] Sursa datei: huggingface-createdqwen3.5-plus Lansat: feb. 2026 Scor: 84.8% CI 95%: [82.3%, 87.4%] Sursa datei: name-suffixqwen3.5-397b-a17b Lansat: feb. 2026 Scor: 86.4% CI 95%: [83.9%, 88.8%] Sursa datei: huggingface-createdqwen3.5-flash Lansat: feb. 2026 Scor: 82.3% CI 95%: [79.6%, 85.0%] Sursa datei: name-suffixqwen3.5-35b-a3b Lansat: feb. 2026 Scor: 83.5% CI 95%: [81.3%, 85.6%] Sursa datei: huggingface-createdqwen3.5-9B Lansat: feb. 2026 Scor: 79.0% CI 95%: [76.6%, 81.3%] Sursa datei: huggingface-createdgpt-5.4-pro-2026-03-05 Lansat: mar. 2026 Scor: 94.6% CI 95%: [93.0%, 96.2%] Sursa datei: name-suffixgpt-5.4-progemma-4-31b-it Lansat: mar. 2026 Scor: 75.8% CI 95%: [72.7%, 78.8%] Sursa datei: huggingface-createdgemma-4-26b-a4b-it Lansat: mar. 2026 Scor: 73.2% CI 95%: [70.1%, 76.4%] Sursa datei: huggingface-createdglm-5.1 Lansat: apr. 2026 Scor: 89.9% CI 95%: [87.8%, 92.0%] Sursa datei: huggingface-createdkimi-k2.6 Lansat: apr. 2026 Scor: 90.8% CI 95%: [89.1%, 92.5%] Sursa datei: huggingface-createdqwen3.6-35b-a3b Lansat: apr. 2026 Scor: 84.8% CI 95%: [82.3%, 87.4%] Sursa datei: huggingface-createdqwen3.6-27b Lansat: apr. 2026 Scor: 85.9% CI 95%: [83.4%, 88.3%] Sursa datei: huggingface-createdqwen3.7-max Lansat: mai 2026 Scor: 90.9% CI 95%: [88.9%, 93.0%] Sursa datei: name-suffixMiniMax-M3 Lansat: iun. 2026 Scor: 90.9% CI 95%: [88.9%, 93.0%] Sursa datei: huggingface-createdkimi-k2.7-code Lansat: iun. 2026 Scor: 87.9% CI 95%: [85.6%, 90.2%] Sursa datei: huggingface-createdkimi-k3 Lansat: iun. 2026 Scor: 93.1% CI 95%: [91.6%, 94.6%] Sursa datei: huggingface-createdglm-5.2 Lansat: iun. 2026 Scor: 91.9% CI 95%: [90.2%, 93.5%] Sursa datei: huggingface-createdInkling Lansat: iul. 2026 Scor: 88.3% CI 95%: [86.3%, 90.2%] Sursa datei: huggingface-createdinkling-small Lansat: iul. 2026 Scor: 88.5% CI 95%: [86.6%, 90.4%] Sursa datei: huggingface-createddeepseek-v4-flash-0731 Lansat: iul. 2026 Scor: 91.0% CI 95%: [89.3%, 92.8%] Sursa datei: huggingface-createddeepseek-v4-pro Lansat: aug. 2026 Scor: 90.9% CI 95%: [88.9%, 93.0%] Sursa datei: name-suffixdeepseek-v4-pro-0813 Lansat: aug. 2026 Scor: 91.7% CI 95%: [90.1%, 93.2%] Sursa datei: huggingface-createdglm-5.3 Lansat: aug. 2026 Scor: 90.9% CI 95%: [89.3%, 92.5%] Sursa datei: huggingface-createdglm-5.3-flash Lansat: aug. 2026 Scor: 90.2% CI 95%: [88.4%, 91.9%] Sursa datei: huggingface-createdgpt-5.6-luna Lansat: sept. 2026 Scor: 91.6% CI 95%: [89.9%, 93.3%] Sursa datei: lmarena-first-seengpt-5.6-sol Lansat: sept. 2026 Scor: 93.5% CI 95%: [91.9%, 95.1%] Sursa datei: lmarena-first-seengpt-5.6-terra Lansat: sept. 2026 Scor: 93.3% CI 95%: [91.8%, 94.8%] Sursa datei: lmarena-first-seengpt-5.4-2026-03-05 Lansat: sept. 2026 Scor: 93.3% CI 95%: [91.5%, 95.1%] Sursa datei: lmarena-first-seengemini-3.1-pro-preview Lansat: sept. 2026 Scor: 94.4% CI 95%: [92.8%, 96.1%] Sursa datei: lmarena-first-seenclaude-sonnet-4-6 Lansat: sept. 2026 Scor: 87.4% CI 95%: [85.4%, 89.4%] Sursa datei: lmarena-first-seengrok-4.6 Lansat: sept. 2026 Scor: 94.0% CI 95%: [92.6%, 95.5%] Sursa datei: lmarena-first-seengemini-3.6-flash Lansat: sept. 2026 Scor: 94.1% CI 95%: [92.7%, 95.5%] Sursa datei: lmarena-first-seenclaude-opus-4-7 Lansat: sept. 2026 Scor: 90.2% CI 95%: [88.3%, 92.0%] Sursa datei: lmarena-first-seenclaude-opus-4-6 Lansat: sept. 2026 Scor: 90.5% CI 95%: [88.8%, 92.2%] Sursa datei: lmarena-first-seenclaude-opus-5 Lansat: sept. 2026 Scor: 93.9% CI 95%: [92.4%, 95.4%] Sursa datei: lmarena-first-seengpt-5.5 Lansat: sept. 2026 Scor: 90.7% CI 95%: [88.9%, 92.4%] Sursa datei: lmarena-first-seenclaude-fable-5 Lansat: sept. 2026 Scor: 85.9% CI 95%: [83.4%, 88.3%] Sursa datei: lmarena-first-seengpt-6-astra Lansat: sept. 2026 Scor: 95.8% CI 95%: [94.4%, 97.1%] Sursa datei: lmarena-first-seengpt-6-astraclaude-opus-4-8 Lansat: sept. 2026 Scor: 91.0% CI 95%: [89.1%, 93.0%] Sursa datei: lmarena-first-seengrok-4.5 Lansat: sept. 2026 Scor: 93.4% CI 95%: [92.0%, 94.9%] Sursa datei: lmarena-first-seenmuse-spark Lansat: sept. 2026 Scor: 89.8% CI 95%: [87.6%, 92.0%] Sursa datei: lmarena-first-seenqwen3.7-plus Lansat: sept. 2026 Scor: 87.9% CI 95%: [85.6%, 90.2%] Sursa datei: lmarena-first-seenclaude-sonnet-5 Lansat: sept. 2026 Scor: 90.5% CI 95%: [88.7%, 92.3%] Sursa datei: lmarena-first-seengemini-3.5-flash Lansat: sept. 2026 Scor: 92.8% CI 95%: [91.2%, 94.4%] Sursa datei: lmarena-first-seengpt-5.5-instant Lansat: sept. 2026 Scor: 82.5% CI 95%: [80.3%, 84.8%] Sursa datei: lmarena-first-seengpt-5.4-nano-2026-03-17 Lansat: oct. 2026 Scor: 78.5% CI 95%: [76.0%, 80.9%] Sursa datei: lmarena-first-seengemini-3.5-flash-lite Lansat: oct. 2026 Scor: 83.3% CI 95%: [80.7%, 86.0%] Sursa datei: lmarena-first-seenqwen3.6-plus Lansat: oct. 2026 Scor: 88.4% CI 95%: [86.1%, 90.7%] Sursa datei: lmarena-first-seengpt-6-luna Lansat: oct. 2026 Scor: 90.5% CI 95%: [88.6%, 92.3%] Sursa datei: lmarena-first-seengpt-6-sol Lansat: oct. 2026 Scor: 94.3% CI 95%: [92.8%, 95.8%] Sursa datei: lmarena-first-seenclaude-sonnet-5-5 Lansat: oct. 2026 Scor: 95.6% CI 95%: [94.2%, 97.0%] Sursa datei: lmarena-first-seengpt-6.1-sol Lansat: oct. 2026 Scor: 95.4% CI 95%: [94.0%, 96.8%] Sursa datei: lmarena-first-seengpt-5.4-mini-2026-03-17 Lansat: oct. 2026 Scor: 86.9% CI 95%: [84.5%, 89.3%] Sursa datei: lmarena-first-seengrok-4.7 Lansat: oct. 2026 Scor: 92.7% CI 95%: [91.2%, 94.3%] Sursa datei: lmarena-first-seengrok-4.3 Lansat: oct. 2026 Scor: 88.8% CI 95%: [86.9%, 90.8%] Sursa datei: lmarena-first-seenqwen3.8-max-0902 Lansat: oct. 2026 Scor: 92.3% CI 95%: [90.6%, 94.0%] Sursa datei: lmarena-first-seenclaude-opus-5-5 Lansat: oct. 2026 Scor: 90.6% CI 95%: [88.6%, 92.5%] Sursa datei: lmarena-first-seengemini-3.8-flash Lansat: oct. 2026 Scor: 95.4% CI 95%: [94.0%, 96.8%] Sursa datei: lmarena-first-seenqwen3.8-max Lansat: oct. 2026 Scor: 92.7% CI 95%: [91.0%, 94.4%] Sursa datei: lmarena-first-seengemini-3.7-flash Lansat: oct. 2026 Scor: 94.8% CI 95%: [93.5%, 96.2%] Sursa datei: lmarena-first-seengemini-4-argon Lansat: oct. 2026 Scor: 96.4% CI 95%: [95.2%, 97.6%] Sursa datei: lmarena-first-seengemini-4-argonqwen3.6-max-preview Lansat: oct. 2026 Scor: 87.4% CI 95%: [85.0%, 89.7%] Sursa datei: lmarena-first-seenclaude-haiku-5-5 Lansat: oct. 2026 Scor: 89.6% CI 95%: [87.7%, 91.6%] Sursa datei: lmarena-first-seennemotron-3-ultra Lansat: oct. 2026 Scor: 85.4% CI 95%: [82.8%, 87.9%] Sursa datei: lmarena-first-seendeepseek-v4.1-flash Lansat: oct. 2026 Scor: 89.8% CI 95%: [88.1%, 91.6%] Sursa datei: huggingface-created
Modelul la data lui de lansareCel mai bun scor curent (frontiera)Intervalul de încredere al scorului

Citește-l ca pe un grafic de frontieră, nu ca pe o traiectorie per model. Scorul unui model fix pe un benchmark fix nu se schimbă în timp — numărul GPQA al lui Claude 3.5 Sonnet de azi este același pe care Anthropic l-a raportat la lansare. Linia urcă pentru că apar mereu modele mai bune, nu pentru că vreun model anume s-a îmbunătățit. Ca să vezi cum se schimbă în timp scorul unui singur model (pe măsură ce evaluarea lui este rerulată cu alte scaffold-uri, de exemplu) e nevoie de alt fel de date: instantanee zilnice ale fiecărui clasament, pe care le vom reprezenta după ce se vor acumula câteva săptămâni.

68 de modele nu au o dată de lansare cunoscută și nu sunt reprezentate.

Benchmark-uri · frontieră

Reperele frontierei

De fiecare dată când un plafon s-a mutat: cine, ce benchmark, cu cât. Atinge un sparkline ca să alegi un benchmark — graficul de mai sus îl urmează; atinge o organizație ca să filtrezi după laborator. Fiecare eveniment este diferența celui mai bun scor curent pe care îl desenează graficul.

114 din 114 evenimente de frontieră

  1. oct. 202696.4%

    gemini-4-argon Google DeepMind stabilește GPQA Diamond

    +0.6 pp
  2. oct. 202676.0%

    gemini-4-argon Google DeepMind stabilește SimpleQA Verified

    +0.4 pp
  3. sept. 202690.4%

    claude-opus-5 Anthropic stabilește ARC-AGI-2

    +7.1 pp
  4. sept. 202695.0%

    gpt-6-astra OpenAI stabilește ARC-AGI-2

    +4.6 pp
  5. sept. 202695.8%

    gpt-6-astra OpenAI stabilește GPQA Diamond

    +1.2 pp
  6. sept. 202646.5%

    claude-fable-5-1 Anthropic stabilește Humanity's Last Exam

    +2.2 pp
  7. sept. 202654.8%

    gpt-6-astra OpenAI stabilește Humanity's Last Exam

    +8.3 pp
  8. sept. 202699.2%

    muse-spark-1.3 Meta AI stabilește OTIS Mock AIME

    +0.6 pp
  9. sept. 2026100.0%

    gpt-5.6-sol OpenAI stabilește OTIS Mock AIME

    +0.8 pp
  10. sept. 202675.9%

    grok-4.6 xAI stabilește SimpleBench

    +1.8 pp
  11. sept. 202680.6%

    claude-opus-5 Anthropic stabilește SimpleBench

    +4.7 pp
  12. sept. 202681.9%

    claude-fable-5 Anthropic stabilește SimpleBench

    +1.3 pp
  13. sept. 202663.0%

    gpt-5.5 OpenAI stabilește SimpleQA Verified

    +7.2 pp
  14. sept. 202670.7%

    claude-fable-5 Anthropic stabilește SimpleQA Verified

    +7.7 pp
  15. sept. 202675.6%

    gpt-6-astra OpenAI stabilește SimpleQA Verified

    +4.9 pp
  16. sept. 202679.3%

    gemini-3.5-flash Google DeepMind stabilește SWE-bench Verified

    +0.6 pp
  17. sept. 202683.5%

    claude-opus-4-7 Anthropic stabilește SWE-bench Verified

    +4.1 pp
  18. sept. 202679.8%

    claude-opus-4-6 Anthropic stabilește Terminal-Bench

    +14.9 pp
  19. sept. 202684.7%

    gpt-5.5 OpenAI stabilește Terminal-Bench

    +4.9 pp
  20. aug. 202698.6%

    deepseek-v4-pro-0813 DeepSeek stabilește OTIS Mock AIME

    +1.4 pp