Benchmark Almanac

EQ-Bench 3

Claude Fable 5 2,096 leads since Jun 2026 Fading

Models respond in character in multi-turn role-plays and analyze transcripts of social conflict. A judge model gives rubric scores and pairwise Elo ratings.

EQ-Bench 4 has replaced it, and EQ-Bench 3 is now a legacy leaderboard.

Made by EQ-Bench Released Apr 2025 Multi-turn conversation All tasks public Last in a launch post Oct 2026 , dropped before it saturated Last model added Oct 2026 Scores refreshed 2 Nov 2026
Level
Everyday Most adults, without training
Scored
Graded by a model A language model scores each response, usually against a rubric.
Answers
Open-ended Free-form text, documents, designs or media judged as a whole.
EQ-bench/eqbench3 70 stars Leaderboard eqbench.com
82
Entries on its own leaderboard, 78 models we track
4
Launch posts list it, from 3 labs
2
System cards report it, from 2 labs

Claude Fable 5 leads with 2,096

Best score per entry from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

82 entries. Scroll sideways for the rest.

  1. 2,096 Claude Fable 5
  2. 2,058 Claude Opus 4.8
  3. 1,914 Claude Opus 4.7
  4. 1,712 Claude Opus 4.6
  5. 1,710 Claude Sonnet 4.6
  6. 1,604 HiveLabsAI/hivemind-32b-preview
  7. 1,563 GLM-5.2
  8. 1,555 DeepSeek-V4-Pro
  9. 1,554 GPT-5.5
  10. 1,553 Gemini 3 Pro
  11. 1,552 openrouter/horizon-alpha
  12. 1,548 Kimi K2.6
  13. 1,548 Kimi K2
  14. 1,544 GPT-5.2
  15. 1,539 GLM-5.1
  16. 1,537 GPT-5.4
  17. 1,535 Claude Sonnet 4.5
  18. 1,530 Kimi K2.5
  19. 1,527 Gemini 3.1 Pro
  20. 1,527 GPT-5.1
  21. 1,523 Gemini 2.5 Pro
  22. 1,522 Claude Opus 4.5
  23. 1,511 GPT-5 Chat
  24. 1,504 DeepSeek-V4-Flash
  25. 1,500 o3
  26. 1,495 GLM-5
  27. 1,464 Gemma 4 31B IT
  28. 1,463 GLM-4.7
  29. 1,455 Gemma 4 26B A4B
  30. 1,451 Qwen3.5 397B-A17B
  31. 1,446 GPT-5.3 Instant
  32. 1,440 Claude Opus 4
  33. 1,428 chatgpt-4o-latest-2025-04-25
  34. 1,419 GPT-4o
  35. 1,416 Gemma 4 12B
  36. 1,405 Hermes 4 405B
  37. 1,322 GLM-4.5
  38. 1,304 Claude Sonnet 4
  39. 1,292 DeepSeek-R1
  40. 1,279 o4-mini
  41. 1,273 Gemma 4 E4B
  42. 1,271 Gemini 2.5 Pro
  43. 1,267 Qwen3-235B-A22B
  44. 1,257 GPT-4.1
  45. 1,203 Grok 4.1 Fast
  46. 1,199 Nemotron 3.5 Bolt E4B
  47. 1,185 QwQ-32B
  48. 1,178 Gemini 2.5 Flash
  49. 1,165 Grok 3
  50. 1,154 Grok 4
  51. 1,128 Claude 3.7 Sonnet
  52. 1,124 DeepSeek-V3
  53. 1,118 glm-4.7-flash
  54. 1,105 Gemma 3 27B
  55. 1,102 Claude 3.5 Sonnet
  56. 1,097 GPT-4.1 mini
  57. 1,093 Llama-3.1-Nemotron-Ultra-253B-v1
  58. 1,092 Qwen3.5-4B
  59. 1,091 Mistral Small 3.2
  60. 1,088 Grok-3 mini
  61. 1,073 gpt-oss-120b
  62. 1,070 GPT-4.5
  63. 1,069 Qwen3-32B
  64. 1,050 GPT-4.1 nano
  65. 1,001 Gemma 3 4B
  66. 998 Mistral Large 2
  67. 985 gpt-oss-20b
  68. 958 Gemini 2.0 Flash
  69. 917 Qwen3-30B-A3B
  70. 909 Qwen3-8B
  71. 896 Qwen2.5-72B
  72. 893 Grok 4.20
  73. 865 Llama 4 Maverick
  74. 862 Mistral Small 3
  75. 853 Llama 3.1-405B
  76. 832 Mistral Small 3.1
  77. 728 Nanbeige/Nanbeige4-3B-Thinking-2511
  78. 670 Llama 4 Scout
  79. 653 GPT-4
  80. 636 Gemma 2 9B
  81. 588 Nemotron 3 Nano 4B
  82. 200 Llama 3.2 1B
  1. 1 Claude Fable 5 2,096
  2. 2 Claude Opus 4.8 2,058
  3. 3 Claude Opus 4.7 1,914
  4. 4 Claude Opus 4.6 1,712
  5. 5 Claude Sonnet 4.6 1,710
  6. 6 HiveLabsAI/hivemind-32b-preview 1,604
  7. 7 GLM-5.2 1,563
  8. 8 DeepSeek-V4-Pro 1,555
  9. 9 GPT-5.5 1,554
  10. 10 Gemini 3 Pro 1,553
  11. 11 openrouter/horizon-alpha 1,552
  12. 12 Kimi K2.6 1,548
  13. 13 Kimi K2 Jul 2025 1,548
  14. 14 GPT-5.2 1,544
  15. 15 GLM-5.1 1,539
  16. 16 GPT-5.4 1,537
  17. 17 Claude Sonnet 4.5 1,535
  18. 18 Kimi K2.5 1,530
  19. 19 Gemini 3.1 Pro 1,527
  20. 20 GPT-5.1 1,527
  21. 21 Gemini 2.5 Pro Jun 2025 1,523
  22. 22 Claude Opus 4.5 1,522
  23. 23 GPT-5 Chat 1,511
  24. 24 DeepSeek-V4-Flash 1,504
  25. 25 o3 1,500
  26. 26 GLM-5 1,495
  27. 27 Gemma 4 31B IT 1,464
  28. 28 GLM-4.7 1,463
  29. 29 Gemma 4 26B A4B 1,455
  30. 30 Qwen3.5 397B-A17B 1,451
  31. 31 GPT-5.3 Instant 1,446
  32. 32 Claude Opus 4 1,440
  33. 33 chatgpt-4o-latest-2025-04-25 1,428
  34. 34 GPT-4o Mar 2025 1,419
  35. 35 Gemma 4 12B 1,416
  36. 36 Hermes 4 405B 1,405
  37. 37 GLM-4.5 1,322
  38. 38 Claude Sonnet 4 1,304
  39. 39 DeepSeek-R1 May 2025 1,292
  40. 40 o4-mini 1,279
  41. 41 Gemma 4 E4B 1,273
  42. 42 Gemini 2.5 Pro May 2025 1,271
  43. 43 Qwen3-235B-A22B 1,267
  44. 44 GPT-4.1 1,257
  45. 45 Grok 4.1 Fast 1,203
  46. 46 Nemotron 3.5 Bolt E4B 1,199
  47. 47 QwQ-32B 1,185
  48. 48 Gemini 2.5 Flash Sep 2025 1,178
  49. 49 Grok 3 1,165
  50. 50 Grok 4 1,154
  51. 51 Claude 3.7 Sonnet 1,128
  52. 52 DeepSeek-V3 Mar 2025 1,124
  53. 53 glm-4.7-flash 1,118
  54. 54 Gemma 3 27B 1,105
  55. 55 Claude 3.5 Sonnet 1,102
  56. 56 GPT-4.1 mini 1,097
  57. 57 Llama-3.1-Nemotron-Ultra-253B-v1 1,093
  58. 58 Qwen3.5-4B 1,092
  59. 59 Mistral Small 3.2 1,091
  60. 60 Grok-3 mini 1,088
  61. 61 gpt-oss-120b 1,073
  62. 62 GPT-4.5 1,070
  63. 63 Qwen3-32B 1,069
  64. 64 GPT-4.1 nano 1,050
  65. 65 Gemma 3 4B 1,001
  66. 66 Mistral Large 2 Nov 2024 998
  67. 67 gpt-oss-20b 985
  68. 68 Gemini 2.0 Flash Feb 2025 958
  69. 69 Qwen3-30B-A3B 917
  70. 70 Qwen3-8B 909
  71. 71 Qwen2.5-72B 896
  72. 72 Grok 4.20 893
  73. 73 Llama 4 Maverick 865
  74. 74 Mistral Small 3 862
  75. 75 Llama 3.1-405B 853
  76. 76 Mistral Small 3.1 832
  77. 77 Nanbeige/Nanbeige4-3B-Thinking-2511 728
  78. 78 Llama 4 Scout 670
  79. 79 GPT-4 Mar 2023 653
  80. 80 Gemma 2 9B 636
  81. 81 Nemotron 3 Nano 4B 588
  82. 82 Llama 3.2 1B 200
#EntryScoreSettingDateSource
1 Claude Fable 5 2,096 Jun 2026 Official leaderboard
2 Claude Opus 4.8 2,058 May 2026 Official leaderboard
3 Claude Opus 4.7 1,914 Apr 2026 Official leaderboard
4 Claude Opus 4.6 1,712 Feb 2026 Official leaderboard
5 Claude Sonnet 4.6 1,710 Feb 2026 Official leaderboard
6 HiveLabsAI/hivemind-32b-preview 1,604 Official leaderboard
7 GLM-5.2 1,563 Jun 2026 Official leaderboard
8 DeepSeek-V4-Pro 1,555 Apr 2026 Official leaderboard
9 GPT-5.5 1,554 Apr 2026 Official leaderboard
10 Gemini 3 Pro 1,553 Nov 2025 Official leaderboard
11 openrouter/horizon-alpha 1,552 Official leaderboard
12 Kimi K2.6 1,548 Apr 2026 Official leaderboard
13 Kimi K2 Jul 2025 1,548 Jul 2025 Official leaderboard
14 GPT-5.2 1,544 Dec 2025 Official leaderboard
15 GLM-5.1 1,539 Apr 2026 Official leaderboard
16 GPT-5.4 1,537 Mar 2026 Official leaderboard
17 Claude Sonnet 4.5 1,535 Sep 2025 Official leaderboard
18 Kimi K2.5 1,530 Jan 2026 Official leaderboard
19 Gemini 3.1 Pro 1,527 Feb 2026 Official leaderboard
20 GPT-5.1 1,527 Nov 2025 Official leaderboard
21 Gemini 2.5 Pro Jun 2025 1,523 Jun 2025 Official leaderboard
22 Claude Opus 4.5 1,522 Nov 2025 Official leaderboard
23 GPT-5 Chat 1,511 Aug 2025 Official leaderboard
24 DeepSeek-V4-Flash 1,504 Apr 2026 Official leaderboard
25 o3 1,500 Apr 2025 Official leaderboard
26 GLM-5 1,495 Feb 2026 Official leaderboard
27 Gemma 4 31B IT 1,464 Apr 2026 Official leaderboard
28 GLM-4.7 1,463 Dec 2025 Official leaderboard
29 Gemma 4 26B A4B 1,455 Apr 2026 Official leaderboard
30 Qwen3.5 397B-A17B 1,451 Feb 2026 Official leaderboard
31 GPT-5.3 Instant 1,446 Mar 2026 Official leaderboard
32 Claude Opus 4 1,440 May 2025 Official leaderboard
33 chatgpt-4o-latest-2025-04-25 1,428 Official leaderboard
34 GPT-4o Mar 2025 1,419 Mar 2025 Official leaderboard
35 Gemma 4 12B 1,416 Jun 2026 Official leaderboard
36 Hermes 4 405B 1,405 Aug 2025 Official leaderboard
37 GLM-4.5 1,322 Aug 2025 Official leaderboard
38 Claude Sonnet 4 1,304 May 2025 Official leaderboard
39 DeepSeek-R1 May 2025 1,292 May 2025 Official leaderboard
40 o4-mini 1,279 Apr 2025 Official leaderboard
41 Gemma 4 E4B 1,273 Mar 2026 Official leaderboard
42 Gemini 2.5 Pro May 2025 1,271 May 2025 Official leaderboard
43 Qwen3-235B-A22B 1,267 Apr 2025 Official leaderboard
44 GPT-4.1 1,257 Apr 2025 Official leaderboard
45 Grok 4.1 Fast 1,203 Nov 2025 Official leaderboard
46 Nemotron 3.5 Bolt E4B 1,199 Oct 2026 Official leaderboard
47 QwQ-32B 1,185 Mar 2025 Official leaderboard
48 Gemini 2.5 Flash Sep 2025 1,178 Sep 2025 Official leaderboard
49 Grok 3 1,165 Apr 2025 Official leaderboard
50 Grok 4 1,154 Jul 2025 Official leaderboard
51 Claude 3.7 Sonnet 1,128 Feb 2025 Official leaderboard
52 DeepSeek-V3 Mar 2025 1,124 Mar 2025 Official leaderboard
53 glm-4.7-flash 1,118 Jan 2026 Official leaderboard
54 Gemma 3 27B 1,105 Mar 2025 Official leaderboard
55 Claude 3.5 Sonnet 1,102 Jun 2024 Official leaderboard
56 GPT-4.1 mini 1,097 Apr 2025 Official leaderboard
57 Llama-3.1-Nemotron-Ultra-253B-v1 1,093 Apr 2025 Official leaderboard
58 Qwen3.5-4B 1,092 Feb 2026 Official leaderboard
59 Mistral Small 3.2 1,091 Jun 2025 Official leaderboard
60 Grok-3 mini 1,088 Apr 2025 Official leaderboard
61 gpt-oss-120b 1,073 Aug 2025 Official leaderboard
62 GPT-4.5 1,070 Feb 2025 Official leaderboard
63 Qwen3-32B 1,069 Apr 2025 Official leaderboard
64 GPT-4.1 nano 1,050 Apr 2025 Official leaderboard
65 Gemma 3 4B 1,001 Mar 2025 Official leaderboard
66 Mistral Large 2 Nov 2024 998 Nov 2024 Official leaderboard
67 gpt-oss-20b 985 Aug 2025 Official leaderboard
68 Gemini 2.0 Flash Feb 2025 958 Feb 2025 Official leaderboard
69 Qwen3-30B-A3B 917 Apr 2025 Official leaderboard
70 Qwen3-8B 909 Apr 2025 Official leaderboard
71 Qwen2.5-72B 896 Sep 2024 Official leaderboard
72 Grok 4.20 893 Feb 2026 Official leaderboard
73 Llama 4 Maverick 865 Apr 2025 Official leaderboard
74 Mistral Small 3 862 Jan 2025 Official leaderboard
75 Llama 3.1-405B 853 Jul 2024 Official leaderboard
76 Mistral Small 3.1 832 Mar 2025 Official leaderboard
77 Nanbeige/Nanbeige4-3B-Thinking-2511 728 Official leaderboard
78 Llama 4 Scout 670 Apr 2025 Official leaderboard
79 GPT-4 Mar 2023 653 Mar 2023 Official leaderboard
80 Gemma 2 9B 636 Jun 2024 Official leaderboard
81 Nemotron 3 Nano 4B 588 Mar 2026 Official leaderboard
82 Llama 3.2 1B 200 Sep 2024 Official leaderboard

Labs report scores for 1 model in their cards. We plot every source in the charts below.

9 models have held number one since its release in Apr 2025

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI Google DeepMind Moonshot AI Anthropic
Own leaderboard 9 leaders
Claude Fable 5 2,096

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Reported by the lab in its model card
0 500 1,000 1,500 2,000 2,500 2023 2024 2025 2026 2027 0 500 1,000 1,500 2,000 2,500 2023 2024 2025 2026 2027

Labs headline it less often now, last in Oct 2026

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

Nous Research
xAI
NVIDIA
DateLaunch postScoreAs the post writes it
23 Oct 2026 Nemotron 3.5 Bolt E4B 38.1 EQ-Bench v3 (non-reasoning)
16 Mar 2026 Nemotron 3 Nano 4B 63.2 EQ-Bench3 (reasoning off)
17 Nov 2025 Grok 4.1 1586 EQ-Bench (Emotional Intelligence Benchmark - Elo (Normalized))
26 Aug 2025 Hermes 4 405B 85.4 EQBench3 (reasoning)

2 system cards report it, from 2 labs

Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.

DateDocumentScoreAs the document writes itSection
23 Oct 2026 Nemotron 3.5 Bolt E4B technical report 38.1 EQ-Bench v3, rubric score, released Q4_K_M imatrix checkpoint (Table 5). BF16 39.4 (Table 3) Capabilities
25 Aug 2025 Hermes 4 405B technical report 85.5 EQBench3, reasoning mode (Table 3). Non-reasoning 74.6. v1 printed 85.4 Capabilities

Versions

Replaces EQ-Bench v2 Replaced by EQ-Bench 4

More in Instructions and chat

All 76 in Instructions and chat
  • IFEval In use Models answer about 500 prompts containing verifiable instructions such as word limits or required keywords. Code checks each instruction for strict and loose accuracy. Gemma 4 31B IT 98.9%
  • IFBench In use 58 new out-of-domain verifiable output constraints. Scored by programmatic constraint checks (strict or loose). Nemotron-Cascade 2 30B-A3B 82.9%
  • MultiChallenge In use Multi-turn conversations testing instruction retention, inference memory, versioned editing and self-coherence. Graded by LLM judge with rubrics. GPT-6.1 Sol 82%
  • Multi-IF In use Models follow verifiable instructions over three-turn conversations in eight languages, extending IFEval. Scored by instruction accuracy at each turn. LFM3-1.7B 89.1%
  • AdvancedIF In use Models respond to over 1,600 prompts with complex, multi-turn, and system-level instructions. A judge model checks each response against expert-written rubrics. MAI-Thinking-1 85%
  • IFStruct In use Models generate JSON or YAML that must follow a requested schema and phrasing constraints, without constrained decoding. A response passes only if every structural constraint holds. LFM3-1.7B 93.7%