Benchmark Almanac

BVB

Claude Fable 5.2 76.4 leads since November 2026
Claude Fable 5.2 76.4 leads since November 2026
In use

The official leaderboard hasn’t scored a model released after Claude Fable 5.2 (November 2026). Labs report their own numbers for one newer model: Inkling 2 Mini at 68.4.

Multimodal agents rebuild real-world videos as animated Blender scenes by writing code. The renders are compared with the source video under a shared cost limit.

It had 31 Hugging Face paper upvotes and 120 GitHub stars.

Released
September 2026
Kind
Video understanding
Level
Expert
Scored
Checked automatically
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Video → Code, 3D, Actions
Interaction
Agentic
Environment
Blender
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
Strongest model tested
Claude Fable 5.2 #2 on the capability index
Last in a launch post
November 2026
Last model added
November 2026
Scores refreshed
November 20th, 2026

In the authors’ words

Multimodal agents can create complex videos in software such as Blender by writing code instead of using diffusion models. Yet video understanding benchmarks still evaluate models mainly through question answering. If an agent truly understands a video, it can reconstruct it programmatically. We introduce BVB, Blender-VideoBench, a benchmark that tests this ability by asking agents to reconstruct real-world videos as animated Blender scenes. To ensure fair comparison, each agent programs the reconstruction through a lightweight harness, Mini-BVB, in an identical sandbox under a shared cost limit. The benchmark renders each reconstruction from its animated camera and evaluates it on two axes: (1) Dual VQA measures how many spatiotemporal facts the reconstruction preserves. (2) Latent Similarity measures how closely the reconstruction matches the source video perceptually. Our overall score, a square-root mean, favors balanced performance. We evaluate 51 configurations from 10 model families and analyze semantic retention, perceptual similarity, reasoning effort, and cost. The best model reaches 88.6 Latent Similarity but retains only 53.7% of the spatiotemporal facts from the source video. Additional reasoning improves perceptual similarity but does not close this gap. In a blind study with 15 raters and five configurations, Latent Similarity correlates strongly with human preference. These results show that programmatic reconstruction is a viable test of agentic video understanding, and that semantic retention remains the main challenge.

Abstract of BVB: Benchmarking Agentic Video Understanding via Programmatic Reconstruction in Blender (arXiv:2609.15478, CC0)
50
Entries on its own leaderboard, 49 models we track
1
Launch post lists it, from 1 lab
1
System card reports it, from 1 lab

Claude Fable 5.2 leads with 76.4 Claude Fable 5.2 leads with 76.4

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

Claude Fable 5.2 leads with 76.4, claims included

Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model. Striped bars are claims, and dotted bars are reruns by Epoch AI or another board.

  1. 1 Claude Fable 5.2 xhigh 76.4
  2. 2 GPT-6 Astra xhigh 76.3
  3. 3 GPT-6.1 Sol xhigh 75.4
  4. 4 GPT-6.1 Terra xhigh 75.2
  5. 5 GPT-6 Sol xhigh 73.4
  6. 6 Gemini 4 Argon high 72.7
  7. 7 Claude Opus 5.5 xhigh 72.4
  8. 8 Claude Sonnet 5.5 xhigh 72.3
  9. 9 GPT-5.6 Sol xhigh 71.8
  10. 10 Inkling 2 Small max, Max · RC17 · effort 1, BVB Worm variant 71.6
  11. 11 Grok 4.6 xhigh 70.9
  12. 12 GPT-6 Luna xhigh 70.3
  13. 13 Grok 4.7 xhigh 70.1
  14. 14 DeepSeek V4.1 Flash 1017 max 70.0
  15. 15 Claude Opus 5 high 69.5
  16. 16 Qwen 3.8 Max high 69.4
  17. 17 Glimpse max 68.9
  18. 18 GPT-5.6 Terra high 68.8
  19. 19 GPT-5.5 high 68.5
  20. 20 Claude Haiku 5.5 xhigh 68.4
  21. 21 GLM-5.3-Flash xhigh 67.9
  22. 22 Muse Wonder xhigh 67.4
  23. 23 Muse Spark 1.3 xhigh 67.1
  24. 24 DeepSeek V4.1 Flash max 66.8
  25. 25 GPT-5.6 Luna high 66.7
  26. 26 Grok 4.5 high 66.4
  27. 27 Gemini 3.8 Flash high 65.9
  28. 28 Claude Sonnet 4.6 none 65.3
  29. 29 GPT-5.2 high 65.2
  30. 30 Claude Sonnet 5 none 64.6
  31. 31 Gemini 3.1 Pro high 64.6
  32. 32 Claude Opus 4.6 none 64.1
  33. 33 Claude Opus 4.7 none 63.6
  34. 34 Muse Spark 1.2 xhigh 63.4
  35. 35 Claude Opus 4.8 none 62.5
  36. 36 Gemini 3 Flash high 62.1
  37. 37 GPT-5.4 none 60.5
  38. 38 GPT-5.4 Mini none 60.2
  39. 39 Kimi K2.5 none 59.6
  40. 40 GLM-5V-Turbo none 59.0
  41. 41 Inkling-Small xhigh 59.0
  42. 42 MiniMax-M3 none 58.2
  43. 43 Qwen3.5-122B-A10B none 56.9
  44. 44 Grok 4.3 high 56.6
  45. 45 Qwen3.5 397B-A17B none 56.3
  46. 46 Seed2.0 Mini none 55.9
  47. 47 Qwen 3.6 Plus none 53.9
  48. 48 Nemotron 3.5 Super VL max 52.1
  49. 49 Seed2.0 Lite (0215) none 51.9
  50. 50 Qwen3.5-27B none 50.7

Source: Official BVB leaderboard

Source: Official BVB leaderboard

Sources: Official BVB leaderboard and labs’ model cards

#EntryScoreSettingDateSource
1 Claude Fable 5.2 76.4 xhigh, mini-BVB Nov 2026 Official leaderboard
2 GPT-6 Astra 76.3 xhigh, mini-BVB Sep 2026 Official leaderboard
3 GPT-6.1 Sol 75.4 xhigh, mini-BVB Sep 2026 Official leaderboard
4 GPT-6.1 Terra 75.2 xhigh, mini-BVB Oct 2026 Official leaderboard
5 GPT-6 Sol 73.4 xhigh, mini-BVB Sep 2026 Official leaderboard
6 Gemini 4 Argon 72.7 high, mini-BVB Sep 2026 Official leaderboard
7 Claude Opus 5.5 72.4 xhigh, mini-BVB Sep 2026 Official leaderboard
8 Claude Sonnet 5.5 72.3 xhigh, mini-BVB Sep 2026 Official leaderboard
9 GPT-5.6 Sol 71.8 xhigh, mini-BVB Jul 2026 Official leaderboard
10 Inkling 2 Small 71.6 max, Max · RC17 · effort 1, BVB Worm variant Nov 2026 Official leaderboard
11 Grok 4.6 70.9 xhigh, mini-BVB Aug 2026 Official leaderboard
12 GPT-6 Luna 70.3 xhigh, mini-BVB Sep 2026 Official leaderboard
13 Grok 4.7 70.1 xhigh, mini-BVB Sep 2026 Official leaderboard
14 DeepSeek V4.1 Flash 1017 70.0 max, mini-BVB Oct 2026 Official leaderboard
15 Claude Opus 5 69.5 high, mini-BVB Jul 2026 Official leaderboard
16 Qwen 3.8 Max 69.4 high, mini-BVB Aug 2026 Official leaderboard
17 Glimpse closed weights 68.9 max, mini-BVB Official leaderboard
18 GPT-5.6 Terra 68.8 high, mini-BVB Jul 2026 Official leaderboard
19 GPT-5.5 68.5 high, mini-BVB Apr 2026 Official leaderboard
20 Claude Haiku 5.5 68.4 xhigh, mini-BVB Oct 2026 Official leaderboard
21 GLM-5.3-Flash 67.9 xhigh, mini-BVB Aug 2026 Official leaderboard
22 Muse Wonder 67.4 xhigh, mini-BVB Oct 2026 Official leaderboard
23 Muse Spark 1.3 67.1 xhigh, mini-BVB Sep 2026 Official leaderboard
24 DeepSeek V4.1 Flash 66.8 max, mini-BVB Sep 2026 Official leaderboard
25 GPT-5.6 Luna 66.7 high, mini-BVB Jul 2026 Official leaderboard
26 Grok 4.5 66.4 high, mini-BVB Jul 2026 Official leaderboard
27 Gemini 3.8 Flash 65.9 high, mini-BVB Sep 2026 Official leaderboard
28 Claude Sonnet 4.6 65.3 none, mini-BVB Feb 2026 Official leaderboard
29 GPT-5.2 65.2 high, mini-BVB Dec 2025 Official leaderboard
30 Claude Sonnet 5 64.6 none, mini-BVB Jun 2026 Official leaderboard
31 Gemini 3.1 Pro 64.6 high, mini-BVB Feb 2026 Official leaderboard
32 Claude Opus 4.6 64.1 none, mini-BVB Feb 2026 Official leaderboard
33 Claude Opus 4.7 63.6 none, mini-BVB Apr 2026 Official leaderboard
34 Muse Spark 1.2 63.4 xhigh, mini-BVB Aug 2026 Official leaderboard
35 Claude Opus 4.8 62.5 none, mini-BVB May 2026 Official leaderboard
36 Gemini 3 Flash 62.1 high, mini-BVB Dec 2025 Official leaderboard
37 GPT-5.4 60.5 none, mini-BVB Mar 2026 Official leaderboard
38 GPT-5.4 Mini 60.2 none, mini-BVB Mar 2026 Official leaderboard
39 Kimi K2.5 59.6 none, mini-BVB Jan 2026 Official leaderboard
40 GLM-5V-Turbo 59.0 none, mini-BVB Apr 2026 Official leaderboard
41 Inkling-Small 59.0 xhigh Jul 2026 Official leaderboard
42 MiniMax-M3 58.2 none, mini-BVB Jun 2026 Official leaderboard
43 Qwen3.5-122B-A10B 56.9 none, mini-BVB Feb 2026 Official leaderboard
44 Grok 4.3 56.6 high, mini-BVB Apr 2026 Official leaderboard
45 Qwen3.5 397B-A17B 56.3 none, mini-BVB Feb 2026 Official leaderboard
46 Seed2.0 Mini 55.9 none, mini-BVB Feb 2026 Official leaderboard
47 Qwen 3.6 Plus 53.9 none, mini-BVB Mar 2026 Official leaderboard
48 Nemotron 3.5 Super VL 52.1 max, mini-BVB Oct 2026 Official leaderboard
49 Seed2.0 Lite (0215) 51.9 none, mini-BVB Feb 2026 Official leaderboard
50 Qwen3.5-27B 50.7 none, mini-BVB Feb 2026 Official leaderboard
#ModelScoreSettingDateSource
1 Claude Fable 5.2 76.4 xhigh, mini-BVB Nov 2026 Official leaderboard
2 GPT-6 Astra 76.3 xhigh, mini-BVB Sep 2026 Official leaderboard
3 GPT-6.1 Sol 75.4 xhigh, mini-BVB Sep 2026 Official leaderboard
4 GPT-6.1 Terra 75.2 xhigh, mini-BVB Oct 2026 Official leaderboard
5 GPT-6 Sol 73.4 xhigh, mini-BVB Sep 2026 Official leaderboard
6 Gemini 4 Argon 72.7 high, mini-BVB Sep 2026 Official leaderboard
7 Claude Opus 5.5 72.4 xhigh, mini-BVB Sep 2026 Official leaderboard
8 Claude Sonnet 5.5 72.3 xhigh, mini-BVB Sep 2026 Official leaderboard
9 GPT-5.6 Sol 71.8 xhigh, mini-BVB Jul 2026 Official leaderboard
10 Inkling 2 Small 71.6 max, Max · RC17 · effort 1, BVB Worm variant Nov 2026 Official leaderboard
11 Grok 4.6 70.9 xhigh, mini-BVB Aug 2026 Official leaderboard
12 GPT-6 Luna 70.3 xhigh, mini-BVB Sep 2026 Official leaderboard
13 Grok 4.7 70.1 xhigh, mini-BVB Sep 2026 Official leaderboard
14 DeepSeek V4.1 Flash 1017 70.0 max, mini-BVB Oct 2026 Official leaderboard
15 Claude Opus 5 69.5 high, mini-BVB Jul 2026 Official leaderboard
16 Qwen 3.8 Max 69.4 high, mini-BVB Aug 2026 Official leaderboard
17 GPT-5.6 Terra 68.8 high, mini-BVB Jul 2026 Official leaderboard
18 GPT-5.5 68.5 high, mini-BVB Apr 2026 Official leaderboard
19 Claude Haiku 5.5 68.4 xhigh, mini-BVB Oct 2026 Official leaderboard
20 GLM-5.3-Flash 67.9 xhigh, mini-BVB Aug 2026 Official leaderboard
21 Muse Wonder 67.4 xhigh, mini-BVB Oct 2026 Official leaderboard
22 Muse Spark 1.3 67.1 xhigh, mini-BVB Sep 2026 Official leaderboard
23 DeepSeek V4.1 Flash 66.8 max, mini-BVB Sep 2026 Official leaderboard
24 GPT-5.6 Luna 66.7 high, mini-BVB Jul 2026 Official leaderboard
25 Grok 4.5 66.4 high, mini-BVB Jul 2026 Official leaderboard
26 Gemini 3.8 Flash 65.9 high, mini-BVB Sep 2026 Official leaderboard
27 Claude Sonnet 4.6 65.3 none, mini-BVB Feb 2026 Official leaderboard
28 GPT-5.2 65.2 high, mini-BVB Dec 2025 Official leaderboard
29 Claude Sonnet 5 64.6 none, mini-BVB Jun 2026 Official leaderboard
30 Gemini 3.1 Pro 64.6 high, mini-BVB Feb 2026 Official leaderboard
31 Claude Opus 4.6 64.1 none, mini-BVB Feb 2026 Official leaderboard
32 Claude Opus 4.7 63.6 none, mini-BVB Apr 2026 Official leaderboard
33 Muse Spark 1.2 63.4 xhigh, mini-BVB Aug 2026 Official leaderboard
34 Claude Opus 4.8 62.5 none, mini-BVB May 2026 Official leaderboard
35 Gemini 3 Flash 62.1 high, mini-BVB Dec 2025 Official leaderboard
36 GPT-5.4 60.5 none, mini-BVB Mar 2026 Official leaderboard
37 GPT-5.4 Mini 60.2 none, mini-BVB Mar 2026 Official leaderboard
38 Kimi K2.5 59.6 none, mini-BVB Jan 2026 Official leaderboard
39 GLM-5V-Turbo 59.0 none, mini-BVB Apr 2026 Official leaderboard
40 Inkling-Small 59.0 xhigh Jul 2026 Official leaderboard
41 MiniMax-M3 58.2 none, mini-BVB Jun 2026 Official leaderboard
42 Qwen3.5-122B-A10B 56.9 none, mini-BVB Feb 2026 Official leaderboard
43 Grok 4.3 56.6 high, mini-BVB Apr 2026 Official leaderboard
44 Qwen3.5 397B-A17B 56.3 none, mini-BVB Feb 2026 Official leaderboard
45 Seed2.0 Mini 55.9 none, mini-BVB Feb 2026 Official leaderboard
46 Qwen 3.6 Plus 53.9 none, mini-BVB Mar 2026 Official leaderboard
47 Nemotron 3.5 Super VL 52.1 max, mini-BVB Oct 2026 Official leaderboard
48 Seed2.0 Lite (0215) 51.9 none, mini-BVB Feb 2026 Official leaderboard
49 Qwen3.5-27B 50.7 none, mini-BVB Feb 2026 Official leaderboard
#ModelScoreSettingDateSource
1 Claude Fable 5.2 76.4 xhigh, mini-BVB Nov 2026 Official leaderboard
2 GPT-6 Astra 76.3 xhigh, mini-BVB Sep 2026 Official leaderboard
3 GPT-6.1 Sol 75.4 xhigh, mini-BVB Sep 2026 Official leaderboard
4 GPT-6.1 Terra 75.2 xhigh, mini-BVB Oct 2026 Official leaderboard
5 GPT-6 Sol 73.4 xhigh, mini-BVB Sep 2026 Official leaderboard
6 Gemini 4 Argon 72.7 high, mini-BVB Sep 2026 Official leaderboard
7 Claude Opus 5.5 72.4 xhigh, mini-BVB Sep 2026 Official leaderboard
8 Claude Sonnet 5.5 72.3 xhigh, mini-BVB Sep 2026 Official leaderboard
9 GPT-5.6 Sol 71.8 xhigh, mini-BVB Jul 2026 Official leaderboard
10 Inkling 2 Small 71.6 max, Max · RC17 · effort 1, BVB Worm variant Nov 2026 Official leaderboard
11 Grok 4.6 70.9 xhigh, mini-BVB Aug 2026 Official leaderboard
12 GPT-6 Luna 70.3 xhigh, mini-BVB Sep 2026 Official leaderboard
13 Grok 4.7 70.1 xhigh, mini-BVB Sep 2026 Official leaderboard
14 DeepSeek V4.1 Flash 1017 70.0 max, mini-BVB Oct 2026 Official leaderboard
15 Claude Opus 5 69.5 high, mini-BVB Jul 2026 Official leaderboard
16 Qwen 3.8 Max 69.4 high, mini-BVB Aug 2026 Official leaderboard
17 GPT-5.6 Terra 68.8 high, mini-BVB Jul 2026 Official leaderboard
18 GPT-5.5 68.5 high, mini-BVB Apr 2026 Official leaderboard
19 Inkling 2 Mini 68.4 Nov 2026 Lab model card
20 Claude Haiku 5.5 68.4 xhigh, mini-BVB Oct 2026 Official leaderboard
21 GLM-5.3-Flash 67.9 xhigh, mini-BVB Aug 2026 Official leaderboard
22 Muse Wonder 67.4 xhigh, mini-BVB Oct 2026 Official leaderboard
23 Muse Spark 1.3 67.1 xhigh, mini-BVB Sep 2026 Official leaderboard
24 DeepSeek V4.1 Flash 66.8 max, mini-BVB Sep 2026 Official leaderboard
25 GPT-5.6 Luna 66.7 high, mini-BVB Jul 2026 Official leaderboard
26 Grok 4.5 66.4 high, mini-BVB Jul 2026 Official leaderboard
27 Gemini 3.8 Flash 65.9 high, mini-BVB Sep 2026 Official leaderboard
28 Claude Sonnet 4.6 65.3 none, mini-BVB Feb 2026 Official leaderboard
29 GPT-5.2 65.2 high, mini-BVB Dec 2025 Official leaderboard
30 Claude Sonnet 5 64.6 none, mini-BVB Jun 2026 Official leaderboard
31 Gemini 3.1 Pro 64.6 high, mini-BVB Feb 2026 Official leaderboard
32 Claude Opus 4.6 64.1 none, mini-BVB Feb 2026 Official leaderboard
33 Claude Opus 4.7 63.6 none, mini-BVB Apr 2026 Official leaderboard
34 Muse Spark 1.2 63.4 xhigh, mini-BVB Aug 2026 Official leaderboard
35 Claude Opus 4.8 62.5 none, mini-BVB May 2026 Official leaderboard
36 Gemini 3 Flash 62.1 high, mini-BVB Dec 2025 Official leaderboard
37 GPT-5.4 60.5 none, mini-BVB Mar 2026 Official leaderboard
38 GPT-5.4 Mini 60.2 none, mini-BVB Mar 2026 Official leaderboard
39 Kimi K2.5 59.6 none, mini-BVB Jan 2026 Official leaderboard
40 GLM-5V-Turbo 59.0 none, mini-BVB Apr 2026 Official leaderboard
41 Inkling-Small 59.0 xhigh Jul 2026 Official leaderboard
42 MiniMax-M3 58.2 none, mini-BVB Jun 2026 Official leaderboard
43 Qwen3.5-122B-A10B 56.9 none, mini-BVB Feb 2026 Official leaderboard
44 Grok 4.3 56.6 high, mini-BVB Apr 2026 Official leaderboard
45 Qwen3.5 397B-A17B 56.3 none, mini-BVB Feb 2026 Official leaderboard
46 Seed2.0 Mini 55.9 none, mini-BVB Feb 2026 Official leaderboard
47 Qwen 3.6 Plus 53.9 none, mini-BVB Mar 2026 Official leaderboard
48 Nemotron 3.5 Super VL 52.1 max, mini-BVB Oct 2026 Official leaderboard
49 Seed2.0 Lite (0215) 51.9 none, mini-BVB Feb 2026 Official leaderboard
50 Qwen3.5-27B 50.7 none, mini-BVB Feb 2026 Official leaderboard

Labs report scores for 1 model in their cards. Switch the scores above to rank them together.

Rankings by part

We read 12 parts from the benchmark’s own leaderboard in 2 groups: dimension and task type. Each ranks models on one part of the test. Claude Fable 5.2 leads 4 of them, and 4 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Dual VQA

49 models we track have a score here.

  1. 1 Claude Fable 5.2 66.5
  2. 2 GPT-6 Astra 64.7
  3. 3 GPT-6.1 Terra ↑1 63.1
  4. 4 GPT-6.1 Sol ↓1 62.4
  5. 5 GPT-6 Sol 62.2
  6. 6 Claude Sonnet 5.5 ↑2 61.5
  7. 7 Claude Opus 5.5 61.3
  8. 8 Gemini 4 Argon ↓2 61.2
  9. 9 Inkling 2 Small ↑1 59.9
  10. 10 GPT-5.6 Sol ↓1 59.4

BVB score versus cost per scene

The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.

50 60 70 80 $0$1$2$3 Cost per scene ($), lower is better Overall score (score) Claude Fable 5.2 (xhigh)GPT-6 Astra (xhigh)GPT-6.1 Sol (xhigh)GPT-6.1 Terra (xhigh)Inkling 2 Small (max, Max · RC17 · effort 1, BVB Worm variant)GPT-6 Luna (xhigh) 50 60 70 80 $0$1$2$3 Cost per scene ($), lower is better Overall score (score) Claude Fable 5.2 (xhigh)GPT-6 Astra (xhigh)GPT-6.1 Sol (xhigh)GPT-6.1 Terra (xhigh)Inkling 2 Small (max, Max · RC17 · effort 1, BVB Worm variant)GPT-6 Luna (xhigh)
ModelSettingOverall scoreCost per scene
Claude Fable 5.2 xhigh, mini-BVB 76.44 score $2.61
GPT-6 Astra xhigh, mini-BVB 76.33 score $1.72
GPT-6.1 Sol xhigh, mini-BVB 75.41 score $0.5
GPT-6.1 Terra xhigh, mini-BVB 75.16 score $0.3
GPT-6 Astra high, mini-BVB 73.95 score $1.26
GPT-6 Sol xhigh, mini-BVB 73.41 score $0.4
Gemini 4 Argon high, mini-BVB 72.69 score $0.31
Claude Opus 5.5 xhigh, mini-BVB 72.39 score $1.95
Claude Sonnet 5.5 xhigh, mini-BVB 72.27 score $1.18
GPT-5.6 Sol xhigh, mini-BVB 71.78 score $0.78
Inkling 2 Small max, Max · RC17 · effort 1, BVB Worm variant 71.56 score $0.26
Grok 4.6 xhigh, mini-BVB 70.86 score $1.68
GPT-6 Luna xhigh, mini-BVB 70.32 score $0.01
Grok 4.7 xhigh, mini-BVB 70.07 score $1.48
DeepSeek V4.1 Flash 1017 max, mini-BVB 70.02 score $0.04
Claude Opus 5 high, mini-BVB 69.53 score $2.16
Qwen 3.8 Max high, mini-BVB 69.39 score $0.6
GPT-5.6 Sol high, mini-BVB 69.21 score $0.42
Glimpse max, mini-BVB 68.86 score $0.37
GPT-5.6 Terra high, mini-BVB 68.76 score $0.12
GPT-5.5 high, mini-BVB 68.55 score $0.96
Claude Haiku 5.5 xhigh, mini-BVB 68.37 score $0.03
GPT-5.6 Terra xhigh, mini-BVB 68.37 score $0.23
Claude Opus 5.5 high, mini-BVB 68.15 score $0.49
GPT-5.6 Sol none, mini-BVB 68.08 score $0.21
GLM-5.3-Flash xhigh, mini-BVB 67.91 score $0.02
GPT-5.6 Sol medium, mini-BVB 67.9 score $0.27
Muse Wonder xhigh, mini-BVB 67.42 score $1.33
Muse Spark 1.3 xhigh, mini-BVB 67.13 score $0.23
DeepSeek V4.1 Flash max, mini-BVB 66.82 score $0.02
GPT-5.5 medium, mini-BVB 66.7 score $0.56
GPT-5.6 Luna high, mini-BVB 66.69 score $0.08
Grok 4.5 high, mini-BVB 66.42 score $0.3
GPT-5.6 Sol low, mini-BVB 66.28 score $0.2
GPT-5.6 Terra medium, mini-BVB 66.27 score $0.15
Gemini 3.8 Flash high, mini-BVB 65.92 score $0.63
GPT-5.5 low, mini-BVB 65.63 score $0.32
GPT-5.6 Luna medium, mini-BVB 65.59 score $0.05
Claude Sonnet 4.6 none, mini-BVB 65.29 score $0.43
GPT-5.2 high, mini-BVB 65.18 score $0.58
GPT-5.6 Terra low, mini-BVB 64.67 score $0.15
Claude Sonnet 5 none, mini-BVB 64.64 score $0.64
Gemini 3.1 Pro high, mini-BVB 64.62 score $0.67
Claude Sonnet 5 high, mini-BVB 64.16 score $0.42
Claude Opus 4.6 none, mini-BVB 64.09 score $0.67
Claude Sonnet 4.6 high, mini-BVB 64.05 score $0.75
Claude Opus 4.7 none, mini-BVB 63.57 score $0.37
Muse Spark 1.2 xhigh, mini-BVB 63.4 score $0.24
GPT-5.6 Luna low, mini-BVB 63.24 score $0.03
Claude Opus 4.8 none, mini-BVB 62.55 score $0.32
Claude Opus 4.8 high, mini-BVB 62.21 score $0.46
Gemini 3 Flash high, mini-BVB 62.14 score $0.08
Claude Opus 4.7 high, mini-BVB 62.08 score $0.35
GPT-5.2 none, mini-BVB 61.87 score $0.12
GPT-5.4 none, mini-BVB 60.53 score $0.2
GPT-5.5 none, mini-BVB 60.34 score $0.16
GPT-5.4 Mini none, mini-BVB 60.16 score $0.03
Kimi K2.5 none, mini-BVB 59.64 score $0.15
GLM-5V-Turbo none, mini-BVB 59 score $0.28
MiniMax-M3 none, mini-BVB 58.18 score $0.06
Qwen3.5-122B-A10B none, mini-BVB 56.85 score $0.23
Grok 4.3 high, mini-BVB 56.58 score $0.1
Qwen3.5 397B-A17B none, mini-BVB 56.32 score $0.18
Seed2.0 Mini none, mini-BVB 55.87 score $0.03
Qwen 3.6 Plus none, mini-BVB 53.94 score $0.17
Grok 4.3 none, mini-BVB 52.37 score $0.05
Nemotron 3.5 Super VL max, mini-BVB 52.11 score $0.12
Seed2.0 Lite (0215) none, mini-BVB 51.9 score $0.04
Qwen3.5-27B none, mini-BVB 50.74 score $0.07

BVB score versus cost per scene

The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.

50 60 70 80 $0$1$2$3 Cost per scene ($), lower is better Overall score (score) Claude Fable 5.2 (xhigh)GPT-6 Astra (xhigh)GPT-6.1 Sol (xhigh)GPT-6.1 Terra (xhigh)Inkling 2 Small (max, Max · RC17 · effort 1, BVB Worm variant)GPT-6 Luna (xhigh) 50 60 70 80 $0$1$2$3 Cost per scene ($), lower is better Overall score (score) Claude Fable 5.2 (xhigh)GPT-6 Astra (xhigh)GPT-6.1 Sol (xhigh)GPT-6.1 Terra (xhigh)Inkling 2 Small (max, Max · RC17 · effort 1, BVB Worm variant)GPT-6 Luna (xhigh)
ModelSettingOverall scoreCost per scene
Claude Fable 5.2 xhigh, mini-BVB 76.44 score $2.61
GPT-6 Astra xhigh, mini-BVB 76.33 score $1.72
GPT-6.1 Sol xhigh, mini-BVB 75.41 score $0.5
GPT-6.1 Terra xhigh, mini-BVB 75.16 score $0.3
GPT-6 Astra high, mini-BVB 73.95 score $1.26
GPT-6 Sol xhigh, mini-BVB 73.41 score $0.4
Gemini 4 Argon high, mini-BVB 72.69 score $0.31
Claude Opus 5.5 xhigh, mini-BVB 72.39 score $1.95
Claude Sonnet 5.5 xhigh, mini-BVB 72.27 score $1.18
GPT-5.6 Sol xhigh, mini-BVB 71.78 score $0.78
Inkling 2 Small max, Max · RC17 · effort 1, BVB Worm variant 71.56 score $0.26
Grok 4.6 xhigh, mini-BVB 70.86 score $1.68
GPT-6 Luna xhigh, mini-BVB 70.32 score $0.01
Grok 4.7 xhigh, mini-BVB 70.07 score $1.48
DeepSeek V4.1 Flash 1017 max, mini-BVB 70.02 score $0.04
Claude Opus 5 high, mini-BVB 69.53 score $2.16
Qwen 3.8 Max high, mini-BVB 69.39 score $0.6
GPT-5.6 Sol high, mini-BVB 69.21 score $0.42
GPT-5.6 Terra high, mini-BVB 68.76 score $0.12
GPT-5.5 high, mini-BVB 68.55 score $0.96
Claude Haiku 5.5 xhigh, mini-BVB 68.37 score $0.03
GPT-5.6 Terra xhigh, mini-BVB 68.37 score $0.23
Claude Opus 5.5 high, mini-BVB 68.15 score $0.49
GPT-5.6 Sol none, mini-BVB 68.08 score $0.21
GLM-5.3-Flash xhigh, mini-BVB 67.91 score $0.02
GPT-5.6 Sol medium, mini-BVB 67.9 score $0.27
Muse Wonder xhigh, mini-BVB 67.42 score $1.33
Muse Spark 1.3 xhigh, mini-BVB 67.13 score $0.23
DeepSeek V4.1 Flash max, mini-BVB 66.82 score $0.02
GPT-5.5 medium, mini-BVB 66.7 score $0.56
GPT-5.6 Luna high, mini-BVB 66.69 score $0.08
Grok 4.5 high, mini-BVB 66.42 score $0.3
GPT-5.6 Sol low, mini-BVB 66.28 score $0.2
GPT-5.6 Terra medium, mini-BVB 66.27 score $0.15
Gemini 3.8 Flash high, mini-BVB 65.92 score $0.63
GPT-5.5 low, mini-BVB 65.63 score $0.32
GPT-5.6 Luna medium, mini-BVB 65.59 score $0.05
Claude Sonnet 4.6 none, mini-BVB 65.29 score $0.43
GPT-5.2 high, mini-BVB 65.18 score $0.58
GPT-5.6 Terra low, mini-BVB 64.67 score $0.15
Claude Sonnet 5 none, mini-BVB 64.64 score $0.64
Gemini 3.1 Pro high, mini-BVB 64.62 score $0.67
Claude Sonnet 5 high, mini-BVB 64.16 score $0.42
Claude Opus 4.6 none, mini-BVB 64.09 score $0.67
Claude Sonnet 4.6 high, mini-BVB 64.05 score $0.75
Claude Opus 4.7 none, mini-BVB 63.57 score $0.37
Muse Spark 1.2 xhigh, mini-BVB 63.4 score $0.24
GPT-5.6 Luna low, mini-BVB 63.24 score $0.03
Claude Opus 4.8 none, mini-BVB 62.55 score $0.32
Claude Opus 4.8 high, mini-BVB 62.21 score $0.46
Gemini 3 Flash high, mini-BVB 62.14 score $0.08
Claude Opus 4.7 high, mini-BVB 62.08 score $0.35
GPT-5.2 none, mini-BVB 61.87 score $0.12
GPT-5.4 none, mini-BVB 60.53 score $0.2
GPT-5.5 none, mini-BVB 60.34 score $0.16
GPT-5.4 Mini none, mini-BVB 60.16 score $0.03
Kimi K2.5 none, mini-BVB 59.64 score $0.15
GLM-5V-Turbo none, mini-BVB 59 score $0.28
MiniMax-M3 none, mini-BVB 58.18 score $0.06
Qwen3.5-122B-A10B none, mini-BVB 56.85 score $0.23
Grok 4.3 high, mini-BVB 56.58 score $0.1
Qwen3.5 397B-A17B none, mini-BVB 56.32 score $0.18
Seed2.0 Mini none, mini-BVB 55.87 score $0.03
Qwen 3.6 Plus none, mini-BVB 53.94 score $0.17
Grok 4.3 none, mini-BVB 52.37 score $0.05
Nemotron 3.5 Super VL max, mini-BVB 52.11 score $0.12
Seed2.0 Lite (0215) none, mini-BVB 51.9 score $0.04
Qwen3.5-27B none, mini-BVB 50.74 score $0.07

The best score, 76.4, costs $2.61 a task

Runs on the cost frontier score higher than any run that's cheaper. GPT-6.1 Terra at xhigh comes within 1.3 of the best for $0.30 a task, 9 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

OpenAI Thinking Machines Anthropic Cost frontier One model at each effort level
0.00 20.0 40.0 60.0 80.0 $0.01 $0.10 $1.00 $10 Cost per task, log scale Claude Fable 5.2, xhigh GPT-6.1 Sol, xhigh Inkling 2 Small, max, Max · RC17 · effort 1, BVB Worm variant 0.00 20.0 40.0 60.0 80.0 $0.01 $0.10 $1.00 $10 Cost per task, log scale Claude Fable 5.2, xhigh Inkling 2 Small, max, Max · RC17 · effort 1, BVB Worm variant

Source: Official BVB leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Fable 5.2 xhigh mini-BVB 76.4 $2.61
GPT-6 Astra xhigh mini-BVB 76.3 $1.72
GPT-6.1 Sol xhigh mini-BVB 75.4 $0.50
GPT-6.1 Terra xhigh mini-BVB 75.2 $0.30
GPT-6 Astra high mini-BVB 74.0 $1.26
GPT-6 Sol xhigh mini-BVB 73.4 $0.40
Gemini 4 Argon high mini-BVB 72.7 $0.31
Claude Opus 5.5 xhigh mini-BVB 72.4 $1.95
Claude Sonnet 5.5 xhigh mini-BVB 72.3 $1.18
GPT-5.6 Sol xhigh mini-BVB 71.8 $0.78
Inkling 2 Small max, Max · RC17 · effort 1, BVB Worm variant 71.6 $0.26
Grok 4.6 xhigh mini-BVB 70.9 $1.68
GPT-6 Luna xhigh mini-BVB 70.3 $0.01
Grok 4.7 xhigh mini-BVB 70.1 $1.48
DeepSeek V4.1 Flash 1017 max mini-BVB 70.0 $0.04
Claude Opus 5 high mini-BVB 69.5 $2.16
Qwen 3.8 Max high mini-BVB 69.4 $0.60
GPT-5.6 Sol high mini-BVB 69.2 $0.42
GPT-5.6 Terra high mini-BVB 68.8 $0.12
GPT-5.5 high mini-BVB 68.5 $0.96
GPT-5.6 Terra xhigh mini-BVB 68.4 $0.23
Claude Haiku 5.5 xhigh mini-BVB 68.4 $0.03
Claude Opus 5.5 high mini-BVB 68.2 $0.49
GPT-5.6 Sol none mini-BVB 68.1 $0.21
GLM-5.3-Flash xhigh mini-BVB 67.9 $0.02
GPT-5.6 Sol medium mini-BVB 67.9 $0.27
Muse Wonder xhigh mini-BVB 67.4 $1.33
Muse Spark 1.3 xhigh mini-BVB 67.1 $0.23
DeepSeek V4.1 Flash max mini-BVB 66.8 $0.02
GPT-5.5 medium mini-BVB 66.7 $0.56
GPT-5.6 Luna high mini-BVB 66.7 $0.08
Grok 4.5 high mini-BVB 66.4 $0.30
GPT-5.6 Sol low mini-BVB 66.3 $0.20
GPT-5.6 Terra medium mini-BVB 66.3 $0.15
Gemini 3.8 Flash high mini-BVB 65.9 $0.63
GPT-5.5 low mini-BVB 65.6 $0.32
GPT-5.6 Luna medium mini-BVB 65.6 $0.05
Claude Sonnet 4.6 none mini-BVB 65.3 $0.43
GPT-5.2 high mini-BVB 65.2 $0.58
GPT-5.6 Terra low mini-BVB 64.7 $0.15
Claude Sonnet 5 none mini-BVB 64.6 $0.64
Gemini 3.1 Pro high mini-BVB 64.6 $0.67
Claude Sonnet 5 high mini-BVB 64.2 $0.42
Claude Opus 4.6 none mini-BVB 64.1 $0.67
Claude Sonnet 4.6 high mini-BVB 64.0 $0.75
Claude Opus 4.7 none mini-BVB 63.6 $0.37
Muse Spark 1.2 xhigh mini-BVB 63.4 $0.23
GPT-5.6 Luna low mini-BVB 63.2 $0.03
Claude Opus 4.8 none mini-BVB 62.5 $0.32
Claude Opus 4.8 high mini-BVB 62.2 $0.46
Gemini 3 Flash high mini-BVB 62.1 $0.08
Claude Opus 4.7 high mini-BVB 62.1 $0.35
GPT-5.2 none mini-BVB 61.9 $0.12
GPT-5.4 none mini-BVB 60.5 $0.20
GPT-5.5 none mini-BVB 60.3 $0.16
GPT-5.4 Mini none mini-BVB 60.2 $0.03
Kimi K2.5 none mini-BVB 59.6 $0.15
GLM-5V-Turbo none mini-BVB 59.0 $0.28
Inkling-Small xhigh 59.0
MiniMax-M3 none mini-BVB 58.2 $0.06
Qwen3.5-122B-A10B none mini-BVB 56.9 $0.23
Grok 4.3 high mini-BVB 56.6 $0.10
Qwen3.5 397B-A17B none mini-BVB 56.3 $0.18
Seed2.0 Mini none mini-BVB 55.9 $0.03
Qwen 3.6 Plus none mini-BVB 53.9 $0.17
Grok 4.3 none mini-BVB 52.4 $0.05
Nemotron 3.5 Super VL max mini-BVB 52.1 $0.12
Seed2.0 Lite (0215) none mini-BVB 51.9 $0.04
Qwen3.5-27B none mini-BVB 50.7 $0.07

The best score, 76.4, costs $2.61 a task

Runs on the cost frontier score higher than any run that's cheaper. GPT-6.1 Terra at xhigh comes within 1.3 of the best for $0.30 a task, 9 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

OpenAI Thinking Machines Anthropic Cost frontier One model at each effort level
0.00 20.0 40.0 60.0 80.0 $0.01 $0.10 $1.00 $10 Cost per task, log scale Claude Fable 5.2, xhigh GPT-6.1 Sol, xhigh Inkling 2 Small, max, Max · RC17 · effort 1, BVB Worm variant 0.00 20.0 40.0 60.0 80.0 $0.01 $0.10 $1.00 $10 Cost per task, log scale Claude Fable 5.2, xhigh Inkling 2 Small, max, Max · RC17 · effort 1, BVB Worm variant

Source: Official BVB leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Fable 5.2 xhigh mini-BVB 76.4 $2.61
GPT-6 Astra xhigh mini-BVB 76.3 $1.72
GPT-6.1 Sol xhigh mini-BVB 75.4 $0.50
GPT-6.1 Terra xhigh mini-BVB 75.2 $0.30
GPT-6 Astra high mini-BVB 74.0 $1.26
GPT-6 Sol xhigh mini-BVB 73.4 $0.40
Gemini 4 Argon high mini-BVB 72.7 $0.31
Claude Opus 5.5 xhigh mini-BVB 72.4 $1.95
Claude Sonnet 5.5 xhigh mini-BVB 72.3 $1.18
GPT-5.6 Sol xhigh mini-BVB 71.8 $0.78
Inkling 2 Small max, Max · RC17 · effort 1, BVB Worm variant 71.6 $0.26
Grok 4.6 xhigh mini-BVB 70.9 $1.68
GPT-6 Luna xhigh mini-BVB 70.3 $0.01
Grok 4.7 xhigh mini-BVB 70.1 $1.48
DeepSeek V4.1 Flash 1017 max mini-BVB 70.0 $0.04
Claude Opus 5 high mini-BVB 69.5 $2.16
Qwen 3.8 Max high mini-BVB 69.4 $0.60
GPT-5.6 Sol high mini-BVB 69.2 $0.42
GPT-5.6 Terra high mini-BVB 68.8 $0.12
GPT-5.5 high mini-BVB 68.5 $0.96
GPT-5.6 Terra xhigh mini-BVB 68.4 $0.23
Claude Haiku 5.5 xhigh mini-BVB 68.4 $0.03
Claude Opus 5.5 high mini-BVB 68.2 $0.49
GPT-5.6 Sol none mini-BVB 68.1 $0.21
GLM-5.3-Flash xhigh mini-BVB 67.9 $0.02
GPT-5.6 Sol medium mini-BVB 67.9 $0.27
Muse Wonder xhigh mini-BVB 67.4 $1.33
Muse Spark 1.3 xhigh mini-BVB 67.1 $0.23
DeepSeek V4.1 Flash max mini-BVB 66.8 $0.02
GPT-5.5 medium mini-BVB 66.7 $0.56
GPT-5.6 Luna high mini-BVB 66.7 $0.08
Grok 4.5 high mini-BVB 66.4 $0.30
GPT-5.6 Sol low mini-BVB 66.3 $0.20
GPT-5.6 Terra medium mini-BVB 66.3 $0.15
Gemini 3.8 Flash high mini-BVB 65.9 $0.63
GPT-5.5 low mini-BVB 65.6 $0.32
GPT-5.6 Luna medium mini-BVB 65.6 $0.05
Claude Sonnet 4.6 none mini-BVB 65.3 $0.43
GPT-5.2 high mini-BVB 65.2 $0.58
GPT-5.6 Terra low mini-BVB 64.7 $0.15
Claude Sonnet 5 none mini-BVB 64.6 $0.64
Gemini 3.1 Pro high mini-BVB 64.6 $0.67
Claude Sonnet 5 high mini-BVB 64.2 $0.42
Claude Opus 4.6 none mini-BVB 64.1 $0.67
Claude Sonnet 4.6 high mini-BVB 64.0 $0.75
Claude Opus 4.7 none mini-BVB 63.6 $0.37
Muse Spark 1.2 xhigh mini-BVB 63.4 $0.23
GPT-5.6 Luna low mini-BVB 63.2 $0.03
Claude Opus 4.8 none mini-BVB 62.5 $0.32
Claude Opus 4.8 high mini-BVB 62.2 $0.46
Gemini 3 Flash high mini-BVB 62.1 $0.08
Claude Opus 4.7 high mini-BVB 62.1 $0.35
GPT-5.2 none mini-BVB 61.9 $0.12
GPT-5.4 none mini-BVB 60.5 $0.20
GPT-5.5 none mini-BVB 60.3 $0.16
GPT-5.4 Mini none mini-BVB 60.2 $0.03
Kimi K2.5 none mini-BVB 59.6 $0.15
GLM-5V-Turbo none mini-BVB 59.0 $0.28
Inkling-Small xhigh 59.0
MiniMax-M3 none mini-BVB 58.2 $0.06
Qwen3.5-122B-A10B none mini-BVB 56.9 $0.23
Grok 4.3 high mini-BVB 56.6 $0.10
Qwen3.5 397B-A17B none mini-BVB 56.3 $0.18
Seed2.0 Mini none mini-BVB 55.9 $0.03
Qwen 3.6 Plus none mini-BVB 53.9 $0.17
Grok 4.3 none mini-BVB 52.4 $0.05
Nemotron 3.5 Super VL max mini-BVB 52.1 $0.12
Seed2.0 Lite (0215) none mini-BVB 51.9 $0.04
Qwen3.5-27B none mini-BVB 50.7 $0.07

3 models have held number one on the official leaderboard since its release in September 2026 3 models have held number one with claims included since its release in September 2026

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

OpenAI Anthropic
Own leaderboard 3 leaders

Source: Official BVB leaderboard

Scores over time

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

Plot with other benchmarks
On the benchmark’s own leaderboard Reported by the lab in its model card
50.0 60.0 70.0 80.0 2026 2027 50.0 60.0 70.0 80.0 2026 2027

Sources: Official BVB leaderboard and labs’ model cards

One lab put it in a launch post in the last six months

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.

Thinking Machines
DateLaunch postScoreAs the post writes it
Nov 20, 2026 Inkling 2 Mini 68.4 BVB

One system card reports it

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
Nov 20, 2026 Inkling 2 Mini model card 68.4 BVB Capabilities

More in Audio and video

All 695 in Audio and video
  • Video-MMMU In use 300 expert-level lecture videos with questions testing perception, comprehension and adaptation of knowledge. Accuracy. Qwen 3.8 Max 88.7%
  • LVBench In use Hours-long video understanding with multiple-choice questions on reasoning, grounding and summarization. Gemini 4 Argon 91.7%
  • MMVU In use 3,000 expert-annotated questions on videos across 27 disciplines. Multiple-choice and open-ended accuracy. Qwen 3.8 Max 82.4%
  • MLVU In use Multi-task long video understanding (minutes to hours) with multiple-choice and generation tasks. Accuracy. Qwen 3.8 Max 90.8%
  • FLEURS In use Models transcribe read speech in 102 languages built from FLoRes-101 sentences. Model cards usually report word error rate, so lower scores are better. Phi-4-multimodal 4.00%
  • VoiceBench In use Spoken instructions with speaker, environment and content variation. Voice assistant responses scored by GPT judge or rules. Qwen3.5-Omni-Plus 93.1%

Data audit

All 10 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (50 entries), launch posts (1 score) and system cards (1 score).
  • Pass Models scored 50 models we track, plus 1 other entry on its leaderboard.
  • Pass Newest score The newest is from November 2026, for Inkling 2 Mini.
  • Pass Frontier models 7 of the top 10 have a score, led by Claude Fable 5.2 (#2).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on yoloytang.me: 69 rows for models we track and 1 other entry at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check