Skip to main content
Radar

Five or fewer confirmed AI changes, with original sources, on mornings when something changed.A free source-linked morning brief.

Start free brief

Vals EMB (EMB)

Evaluating agents on Excel-based financial modeling tasks

Data verified 24 confirmed releases in the last 30 daysStart free brief

How BenchLM shows EMB

BenchLM mirrors the public Vals AI EMB leaderboard captured from https://www.vals.ai/benchmarks/emb and updated by Vals on August 19, 2026. The snapshot preserves overall scores, uncertainty, latency, cost-per-test metadata, and task-level scores where Vals publishes them.

EMB is display only on BenchLM. Vals proprietary or Vals-hosted aggregate views are useful context, but BenchLM does not use them as weighted ranking inputs or as a replacement for benchmark-native source records.

Snapshot

48 Vals rows10 task viewsprivate datasetTasks: Overall, Template, Scratch, LBO, DCFDisplay only

EMB score on EMB — August 19, 2026

We mirror the published emb score view for EMB. Claude Fable 5 leads the public snapshot at 73.67%, followed by Claude Opus 5 (73.56%) and GPT-5.6 Sol (72.34%). We do not use these results to rank models overall.

48 modelsExternal benchmark mirrorsCurrentDisplay onlyUpdated August 19, 2026

EMB score table (48 models)

Score
1
Claude Fable 5Anthropic · Closed
73.67%
2
Claude Opus 5Anthropic · Closed
73.56%
3
GPT-5.6 SolOpenAI · Closedmax reasoning
72.34%
4
Gemini 3.7 FlashGoogle · Closedhigh reasoning
71.33%
5
Claude Opus 4.8Anthropic · Closed
69.37%
6
GPT-5.6 LunaOpenAI · Closedmax reasoning
67.12%
7
Kimi K3Moonshot AI · Closedmax reasoning
66.40%
8
Claude Sonnet 5Anthropic · Closed
66.32%
9
GPT-5.6 TerraOpenAI · Closedmax reasoning
66.20%
10
Gemini 3.6 FlashGoogle · Closedhigh reasoning
65.41%
11
GPT-5.5OpenAI · Closedxhigh reasoning
64.54%
12
Claude Opus 4.7Anthropic · Closed
63.73%
13
Gemini 3.5 FlashGoogle · Closedhigh reasoning
63.55%
14
Grok 4.6xAI · Closedhigh reasoning
62.73%
15
GLM-5.2Z.AI · Open weight
61.53%
16
Claude Sonnet 4.6Anthropic · Closed
60.15%
17
Qwen3.8 MaxAlibaba · Open weight
60.07%
18
Qwen3.8-27BAlibaba · Open weightxhigh reasoning
59.66%
19
Kimi K2.6Moonshot AI · Open weight
57.85%
20
Qwen3.7 MaxAlibaba · Closed
57.04%
21
Muse Spark 1.2Meta · Closedxhigh reasoning
56.98%
22
DeepSeek V4 Flash 0731DeepSeek · Closedhigh reasoning
56.98%
23
Muse Spark 1.1Meta · Closedxhigh reasoning
56.36%
24
MiMo-V2.5-ProXiaomi · Closed
55.23%
25
MiMo-V2.5Xiaomi · Closed
55.09%
26
Grok 4.5xAI · Closedhigh reasoning
52.91%
27
DeepSeek V4 Pro 0813DeepSeek · Closedmax reasoning
52.80%
28
Gemini 3.1 Pro PreviewGooglehigh reasoning
52.62%
29
DeepSeek V4 Pro 0813DeepSeek · Closedmax reasoning
51.62%
30
Qwen3.7 PlusAlibaba · Closed
49.34%
31
MiniMax M3MiniMax · Open weight
47.76%
32
GPT-5.4 miniOpenAI · Closedxhigh reasoning
45.43%
33
GPT-5.4 nanoOpenAI · Closedhigh reasoning
44.75%
34
Gemini 3.5 Flash-LiteGoogle · Closedhigh reasoning
43.23%
35
InklingThinking Machines Lab · Open weight0.99 reasoning
38.36%
36
Qwen3.6 PlusAlibaba · Closed
32.87%
37
Inkling-SmallThinking Machines Lab · Open weight0.99 reasoning
31.76%
39
28.47%
40
MiniMax M2.7MiniMax · Open weight
28.45%
41
Gemini 3 Flash PreviewGooglehigh reasoning
26.02%
43
Claude Haiku 4.5 ThinkingAnthropic · Closed
22.66%
44
Grok 4.3xAI · Closedhigh reasoning
18.39%
45
16.99%
47
Mistral Medium 3.5Mistral AIhigh reasoning
11.19%

The published EMB snapshot places Claude Fable 5 first at 73.67%. The third row is 1.33 points behind. The broader top-10 range is 8.25 points, so many of the published results sit in a relatively narrow band.

48 models have been evaluated on EMB. The benchmark falls in the External benchmark mirrors category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. EMB is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About EMB

Year

2026

Tasks

Excel-based financial modeling tasks

Format

Accuracy score

Difficulty

Professional finance modeling

BenchLM mirrors the public Vals AI EMB leaderboard as display-only external evidence. The captured snapshot preserves overall scores, task-level scores where Vals publishes them, uncertainty, latency, and cost-per-test metadata. It is excluded from BenchLM weighted rankings.

BenchLM freshness & provenance

Version

EMB 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

FAQ

What does EMB measure?

Evaluating agents on Excel-based financial modeling tasks

Which model leads the published EMB snapshot?

Claude Fable 5 currently leads the published EMB snapshot with 73.67% emb score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on EMB?

48 AI models are included in BenchLM's mirrored EMB snapshot, based on the public leaderboard captured on August 19, 2026.

Last updated: August 19, 2026 · mirrored from the public benchmark leaderboard

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 2,000+ readers.

One email each week. Unsubscribe anytime.