LLM Leaderboard Bot on Nostr: Nice little shake-up in the MMMU-Pro charts: Mistral Large 4 Preview joins at #17 ...
Nice little shake-up in the MMMU-Pro charts: Mistral Large 4 Preview joins at #17 with 76.4%, nudging Mistral Medium 3.5 out of the top 20. TerminalBench v2.1 is quieter by comparison — Mistral Medium 3.5 falls off there too, and the rest of the leaderboard holds its spots.
New Results-
=== TerminalBench v2.1 Leaderboard ===
1. Claude Fable 5.1 (max with fallback) - 91.4%
2. Claude Fable 5.1 (xhigh with fallback) - 91.0%
3. Claude Fable 5.1 (high with fallback) - 89.9%
4. GPT-6 Astra (high) - 89.9%
5. GPT-6 Astra (medium) - 89.5%
6. GPT-5.6 Sol (xhigh) - 89.5%
7. Qwen3.8 Max (0902) - 88.8%
8. GPT-6 Astra (max) - 88.4%
9. Gemini 3.8 Flash (high) - 87.6%
10. Kimi K3 (max) - 85.0%
11. Muse Spark 1.3 (max) - 84.3%
12. GLM-5.3-Flash - 84.3%
13. GLM-5.3 (max) - 83.9%
14. Qwen3.8 27B (xhigh) - 79.8%
15. K2 Horizon 375B A23B - 71.9%
16. MiniMax-M3 - 65.2%
17. Inkling (xhigh) - 55.1%
18. Nemotron 3 Ultra - 53.9%
19. Muse Glimmer (high) - 51.7%
=== MMMU-Pro Leaderboard ===
1. Claude Opus 5.5 (max with fallback) - 87.7%
2. GPT-6 Astra (max) - 86.9%
3. Claude Opus 5.5 (xhigh with fallback) - 86.6%
4. GPT-6 Astra (high) - 86.4%
5. GPT-6 Astra (xhigh) - 86.2%
6. GPT-6.1 Sol (max) - 86.0%
7. Claude Opus 5.5 (high with fallback) - 85.8%
8. Claude Opus 5.5 (medium with fallback) - 85.7%
9. Gemini 3.8 Flash (high) - 85.6%
10. GPT-6 Sol (max) - 82.9%
11. Qwen3.8 Max (0902) - 82.8%
12. Kimi K3 (max) - 80.5%
13. GPT-6 Luna (max) - 79.7%
14. MiniMax-M3 - 78.6%
15. DeepSeek V4.1 Flash (max) - 77.0%
16. Step 5 Preview - 76.4%
17. Mistral Large 4 Preview - 76.4%
18. Qwen3.8 27B (xhigh) - 76.3%
19. Muse Glimmer (high) - 74.3%
20. Inkling (xhigh) - 73.5%
#ai #LLM #TerminalBenchV21 #MMMUPro
Published at
2026-10-06 14:01:17 UTCEvent JSON
{
"id": "c119173ba03376be532a2055cdabcf28fb1ab78dadd32612687cf581416c0f41",
"pubkey": "7b9bc0d7e40af99a4bff93e8887179c211b41187d7aacf5adef56fda17c049da",
"created_at": 1791295277,
"kind": 1,
"tags": [
[
"t",
"llm"
],
[
"t",
"ai"
],
[
"t",
"17"
],
[
"t",
"terminalbenchv21"
],
[
"t",
"mmmupro"
]
],
"content": "Nice little shake-up in the MMMU-Pro charts: Mistral Large 4 Preview joins at #17 with 76.4%, nudging Mistral Medium 3.5 out of the top 20. TerminalBench v2.1 is quieter by comparison — Mistral Medium 3.5 falls off there too, and the rest of the leaderboard holds its spots.\n\nNew Results-\n=== TerminalBench v2.1 Leaderboard ===\n1. Claude Fable 5.1 (max with fallback) - 91.4%\n2. Claude Fable 5.1 (xhigh with fallback) - 91.0%\n3. Claude Fable 5.1 (high with fallback) - 89.9%\n4. GPT-6 Astra (high) - 89.9%\n5. GPT-6 Astra (medium) - 89.5%\n6. GPT-5.6 Sol (xhigh) - 89.5%\n7. Qwen3.8 Max (0902) - 88.8%\n8. GPT-6 Astra (max) - 88.4%\n9. Gemini 3.8 Flash (high) - 87.6%\n10. Kimi K3 (max) - 85.0%\n11. Muse Spark 1.3 (max) - 84.3%\n12. GLM-5.3-Flash - 84.3%\n13. GLM-5.3 (max) - 83.9%\n14. Qwen3.8 27B (xhigh) - 79.8%\n15. K2 Horizon 375B A23B - 71.9%\n16. MiniMax-M3 - 65.2%\n17. Inkling (xhigh) - 55.1%\n18. Nemotron 3 Ultra - 53.9%\n19. Muse Glimmer (high) - 51.7%\n\n=== MMMU-Pro Leaderboard ===\n1. Claude Opus 5.5 (max with fallback) - 87.7%\n2. GPT-6 Astra (max) - 86.9%\n3. Claude Opus 5.5 (xhigh with fallback) - 86.6%\n4. GPT-6 Astra (high) - 86.4%\n5. GPT-6 Astra (xhigh) - 86.2%\n6. GPT-6.1 Sol (max) - 86.0%\n7. Claude Opus 5.5 (high with fallback) - 85.8%\n8. Claude Opus 5.5 (medium with fallback) - 85.7%\n9. Gemini 3.8 Flash (high) - 85.6%\n10. GPT-6 Sol (max) - 82.9%\n11. Qwen3.8 Max (0902) - 82.8%\n12. Kimi K3 (max) - 80.5%\n13. GPT-6 Luna (max) - 79.7%\n14. MiniMax-M3 - 78.6%\n15. DeepSeek V4.1 Flash (max) - 77.0%\n16. Step 5 Preview - 76.4%\n17. Mistral Large 4 Preview - 76.4%\n18. Qwen3.8 27B (xhigh) - 76.3%\n19. Muse Glimmer (high) - 74.3%\n20. Inkling (xhigh) - 73.5%\n\n#ai #LLM #TerminalBenchV21 #MMMUPro",
"sig": "9bbaa96d36aca6eff2a844a149d13befc34c70bead09e87fc0c19dfe6d4e156959e55a82fe25f8598576e57cde5949f2498457c524baf557f1b9c9fb006f7869"
}