Kaelum BenchBack to benchmark areas

Benchmark area

Accounting, Reporting & Financial Analysis

Accounting knowledge, HGB and IFRS reporting, financial analysis, and an integrated accounting capstone.

Provisional category results
  • Weighted category scoreKimi K3 · 5 / 5 identical configuration88.8 / 100

Category formula

60% core average + 40% capstone

Each of the four core tests contributes 15%; the capstone contributes 40%. A score is published only when all five tests use the same configuration key. Missing tests are never treated as zero or removed from the denominator.

Coverage
340 tasks · 5 tests
Protocol
Closed-book · German
Access
No tools or external search
Read full methodology and limitations →

Five tests in this area

Each card shows the strongest observed public calibration runs directly. These are configuration-specific single runs, not statistical rank claims.

FMB-CAP-ACC-03

Accounting Capstone

Public calibration

Integrated accounting analysis across a multi-step financial case.

100 tasks · Hybrid · 40% category weight · 10 basic · 40 intermediate · 50 advanced

Top observed calibration scores

Provisional · configuration-specific single runs

9 shown
  1. 1Kimi K395.3 / 100
  2. 2Kimi K2.692.8 / 100
  3. 3Gemini 3.5 Flash82.9 / 100
  4. 4DeepSeek V4 Flash75.4 / 100
  5. 5Qwen3.6-35B-A3B74.4 / 100
Show ranks 6–9 +
  1. 6Grok 4.573.6 / 100
  2. 7NVIDIA Nemotron 3 Ultra40.7 / 100
  3. 8GLM 5.230.8 / 100
  4. 9MiniMax M330.4 / 100

Tested competencies

  • Statement reconstruction across balance sheet, income statement, equity, and cash flow
  • HGB-to-IFRS separation and bridge
  • Accounting adjustments before ratio analysis
  • Earnings quality, leverage, and stress resilience
Open full benchmark and all models →

FMB-KNOW-ACC-01

Accounting Knowledge

Public calibration

Foundational accounting concepts and financial statement mechanics.

60 tasks · Hybrid · 15% category weight · 12 basic · 30 intermediate · 18 advanced

Top observed calibration scores

Provisional · configuration-specific single runs

10 shown
  1. 1GLM 5.296.0 / 100
  2. 2Grok 4.595.9 / 100
  3. 3DeepSeek V4 Flash95.0 / 100
  4. 4Kimi K2.694.7 / 100
  5. 5Qwen3.6-35B-A3B93.3 / 100
Show ranks 6–10 +
  1. 6Claude Sonnet 5 (Max)93.2 / 100
  2. 7Kimi K391.7 / 100
  3. 8Gemini 3.5 Flash90.3 / 100
  4. 9NVIDIA Nemotron 3 Ultra85.1 / 100
  5. 10MiniMax M380.5 / 100

Tested competencies

  • Balance sheet, income statement, equity, and cash flow
  • Journal entries and account logic
  • Accruals and subsequent measurement
  • Reconciliations and error diagnosis
Open full benchmark and all models →

FMB-KNOW-ACC-HGB-02

HGB Reporting

Public calibration

HGB financial reporting and statement presentation.

60 tasks · Deterministic · 15% category weight · 12 basic · 30 intermediate · 18 advanced

Top observed calibration scores

Provisional · configuration-specific single runs

10 shown
  1. 1Gemini 3.5 Flash92.4 / 100
  2. 2Grok 4.590.7 / 100
  3. 3Kimi K2.690.2 / 100
  4. 4Qwen3.6-35B-A3B89.4 / 100
  5. 5MiniMax M389.3 / 100
Show ranks 6–10 +
  1. 6Kimi K388.4 / 100
  2. 7Claude Sonnet 5 (Max)85.6 / 100
  3. 8DeepSeek V4 Flash84.5 / 100
  4. 9GLM 5.284.5 / 100
  5. 10NVIDIA Nemotron 3 Ultra78.3 / 100

Tested competencies

  • Bookkeeping, inventory, and preparation duties
  • Recognition choices and prohibitions
  • Initial and subsequent measurement
  • HGB statement error reconstruction
Open full benchmark and all models →

FMB-KNOW-ACC-IFRS-02

IFRS Reporting

Public calibration

IFRS financial reporting and statement presentation.

60 tasks · Deterministic · 15% category weight · 12 basic · 30 intermediate · 18 advanced

Top observed calibration scores

Provisional · configuration-specific single runs

10 shown
  1. 1Gemini 3.5 Flash91.9 / 100
  2. 2Grok 4.591.4 / 100
  3. 3Kimi K391.4 / 100
  4. 4Kimi K2.691.1 / 100
  5. 5GLM 5.291.0 / 100
Show ranks 6–10 +
  1. 6Claude Sonnet 5 (Max)90.8 / 100
  2. 7Qwen3.6-35B-A3B90.7 / 100
  3. 8DeepSeek V4 Flash90.6 / 100
  4. 9NVIDIA Nemotron 3 Ultra89.7 / 100
  5. 10MiniMax M388.5 / 100

Tested competencies

  • Presentation, recognition, and events
  • Revenue, provisions, and leases
  • Financial instruments
  • Taxes, employee benefits, and foreign currency
Open full benchmark and all models →

FMB-ANAL-FSA-02

Financial Analysis

Public calibration

Financial statement interpretation and performance analysis.

60 tasks · Deterministic · 15% category weight · 12 basic · 30 intermediate · 18 advanced

Top observed calibration scores

Provisional · configuration-specific single runs

10 shown
  1. 1MiniMax M391.9 / 100
  2. 2Kimi K2.687.5 / 100
  3. 3Gemini 3.5 Flash77.8 / 100
  4. 4Qwen3.6-35B-A3B77.1 / 100
  5. 5GLM 5.267.7 / 100
Show ranks 6–10 +
  1. 6Kimi K366.5 / 100
  2. 7NVIDIA Nemotron 3 Ultra61.2 / 100
  3. 8Claude Sonnet 5 (Max)59.7 / 100
  4. 9Grok 4.556.4 / 100
  5. 10DeepSeek V4 Flash52.2 / 100