Vals MortgageTax (MortgageTax)

We show this table for reference; we do not rank on it.

Data verified September 1, 2026

Vals AI benchmark for mortgage and tax document reasoning, including semantic and numerical extraction task views.

Evaluation results published by Vals AI. Published leaderboard aggregates and evaluation settings only. Dataset contents are not included.

MortgageTax score on MortgageTax — September 1, 2026

We mirror the published mortgagetax score view for MortgageTax. Claude Opus 5 leads the public snapshot at 72.06%, followed by Claude Fable 5.1 (70.79%) and Claude Opus 4.7 (70.27%). We do not use these results to rank models overall.

1Closed

Claude Opus 5

Anthropic

72.06%

Overall 76.05Context 1M

2Closed

Claude Fable 5.1

Anthropic

70.79%

Overall 79.21Context 1M

3Closed

Claude Opus 4.7

Anthropic

70.27%

Overall 63.06Context 1M

98 modelsMultimodal & GroundedCurrentDisplay onlyUpdated September 1, 2026

MortgageTax score table (98 models)

MortgageTax score results for MortgageTax
Rank	Model / configuration	Score	Parameters (B)	Open / closed
1	Claude Opus 5Anthropic	72.06%	Not reported	Closed
2	Claude Fable 5.1Anthropic	70.79%	Not reported	Closed
3	Claude Opus 4.7Anthropic	70.27%	Not reported	Closed
4	Claude Sonnet 5Anthropic	70.03%	Not reported	Closed
5	Claude Opus 4.8Anthropic	69.91%	Not reported	Closed
6	Gemini 3.1 Pro PreviewGooglehigh reasoning	69.40%	Not reported	Unknown
7	Gemini 3 Pro PreviewGooglehigh reasoning	69.08%	Not reported	Unknown
8	Claude Fable 5Anthropic	68.92%	Not reported	Closed
9	Gemini 2.5 ProGoogle	68.92%	Not reported	Closed
10	GPT-5.5OpenAIxhigh reasoning	68.76%	Not reported	Closed
11	Gemini 3 Flash PreviewGooglehigh reasoning	68.72%	Not reported	Unknown
12	Gemini 3.5 Flash-LiteGooglehigh reasoning	68.68%	Not reported	Closed
13	Claude 3.7 SonnetAnthropic	68.68%	Not reported	Closed
14	Claude Opus 4.5Anthropic	68.68%	Not reported	Closed
15	Claude Opus 4.6 (Adaptive)Anthropic	68.52%	Not reported	Closed
16	MiniMax M3MiniMax	68.36%	Not reported	Open
17	GPT-5.4OpenAIxhigh reasoning	68.32%	Not reported	Closed
18	Qwen3.6-27BAlibaba	68.28%	Not reported	Open
19	Gemini 3.5 FlashGooglehigh reasoning	68.12%	Not reported	Closed
20	Gemini 3.1 Flash Lite PreviewGooglehigh reasoning	68.04%	Not reported	Unknown
21	Qwen3.6 PlusAlibaba	67.97%	Not reported	Closed
22	Claude Sonnet 4.6Anthropic	67.73%	Not reported	Closed
23	Claude Opus 4.5 ThinkingAnthropic	67.69%	Not reported	Closed
24	Gemini 3.6 FlashGooglehigh reasoning	67.61%	Not reported	Closed
25	Qwen3.5 FlashAlibaba	67.37%	Not reported	Closed
26	GPT-5.6 TerraOpenAIxhigh reasoning	67.33%	Not reported	Closed
27	GPT-5.6 LunaOpenAImax reasoning	67.29%	Not reported	Closed
28	GPT-5.6 SolOpenAImax reasoning	67.29%	Not reported	Closed
29	Gemini 2.5 Pro Exp 03 25Google	67.17%	Not reported	Unknown
30	GPT-5.2OpenAIxhigh reasoning	67.13%	Not reported	Closed
31	GPT-5 miniOpenAIhigh reasoning	66.89%	Not reported	Closed
32	Claude 3.7 Sonnet ThinkingAnthropic	66.85%	Not reported	Closed
33	Gemini 3.7 FlashGooglehigh reasoning	66.65%	Not reported	Closed
34	Kimi K2.5 ThinkingMoonshot AI	66.53%	Not reported	Unknown
35	Kimi K3Moonshot AI	66.34%	Not reported	Open
36	Qwen3.7 PlusAlibaba	66.18%	Not reported	Closed
37	Muse Spark 1.1Metaxhigh reasoning	66.14%	Not reported	Closed
38	GPT-4.1OpenAIhigh reasoning	65.94%	Not reported	Closed
39	Kimi K2.6Moonshot AI	65.82%	Not reported	Open
40	o3OpenAIhigh reasoning	65.70%	Not reported	Closed
41	GPT-4.1 miniOpenAIhigh reasoning	65.50%	Not reported	Closed
42	GPT-5OpenAIhigh reasoning	65.45%	Not reported	Unknown
43	Muse Spark 1.2Metaxhigh reasoning	65.42%	Not reported	Closed
44	Gemini 3.8 FlashGooglehigh reasoning	65.34%	Not reported	Closed
45	Qwen3.8-27BAlibabaxhigh reasoning	64.94%	Not reported	Open
46	o4-miniOpenAIhigh reasoning	64.83%	Not reported	Closed
47	Grok 4.6xAIhigh reasoning	64.19%	Not reported	Closed
48	Claude 3.5 SonnetAnthropic	64.07%	Not reported	Closed
49	Qwen3.8 MaxAlibaba	63.99%	Not reported	Open
50	Claude Sonnet 4.5 ThinkingAnthropic	63.99%	Not reported	Closed
51	InklingThinking Machines Lab0.99 reasoning	63.99%	Not reported	Open
52	GPT-5.4 miniOpenAIxhigh reasoning	63.51%	Not reported	Closed
53	Gemini 2.5 Flash Preview 09 2025Google	62.60%	Not reported	Unknown
54	Claude Sonnet 4Anthropic	62.47%	Not reported	Unknown
55	Inkling-SmallThinking Machines Lab0.99 reasoning	62.28%	Not reported	Open
56	Claude Haiku 4.5 ThinkingAnthropic	62.16%	Not reported	Closed
57	Magistral Small 2509Mistral AI	62.12%	Not reported	Unknown
58	Gemini 2.5 Flash Preview 09 2025 ThinkingGoogle	61.92%	Not reported	Unknown
59	Grok 4.5xAIhigh reasoning	61.80%	Not reported	Closed
60	Gemma 4 31b ItGooglehigh reasoning	61.37%	Not reported	Unknown
61	GPT-5.1OpenAIhigh reasoning	61.37%	Not reported	Closed
62	Mistral Small 2503Mistral AI	61.21%	Not reported	Unknown
63	Claude Opus 4.1Anthropic	61.09%	Not reported	Unknown
64	GPT-4oOpenAIhigh reasoning	60.97%	Not reported	Closed
65	Qwen3.5 Plus ThinkingAlibaba	60.77%	Not reported	Unknown
66	Gemini 2.0 Flash ExpGoogle	60.37%	Not reported	Unknown
67	Gemini 2.0 Flash 001Google	59.66%	Not reported	Unknown
68	MiMo-V2.5Xiaomi	59.26%	Not reported	Closed
69	GPT-5.4 nanoOpenAIhigh reasoning	59.10%	Not reported	Closed
70	Claude Opus 4Anthropic	58.59%	Not reported	Unknown
71	Llama4 Maverick Instruct BasicFireworks AI	58.51%	Not reported	Unknown
72	Gemini 2.5 Flash Preview 04 17Google	58.43%	Not reported	Unknown
73	Meta Llama Llama 4 Scout 17B 16E InstructTogether AI	57.75%	Not reported	Unknown
74	Gemini 2.5 Flash Lite Preview 09 2025 ThinkingGoogle	57.55%	Not reported	Unknown
75	Gemini 2.5 Flash Preview 04 17 ThinkingGoogle	57.51%	Not reported	Unknown
76	GPT-4oOpenAIhigh reasoning	57.43%	Not reported	Closed
77	Gemini 1.5 Pro 002Google	56.76%	Not reported	Unknown
78	Claude Opus 4.1 20250805 ThinkingAnthropic	56.12%	Not reported	Unknown
79	Gemini 2.5 Flash Lite Preview 09 2025Google	55.80%	Not reported	Unknown
80	GPT-4o miniOpenAIhigh reasoning	54.49%	Not reported	Closed
81	GPT-5 nanoOpenAIhigh reasoning	53.62%	Not reported	Closed
82	GPT-4.1 nanoOpenAIhigh reasoning	52.82%	Not reported	Closed
83	Mistral Large 2512Mistral AI	52.11%	Not reported	Unknown
84	Magistral Medium 2509Mistral AI	51.87%	Not reported	Unknown
85	Claude Sonnet 4 20250514 ThinkingAnthropic	49.88%	Not reported	Unknown
86	Grok 4.3xAI	48.25%	Not reported	Closed
87	Grok 4.20 0309 ReasoningxAI	45.35%	Not reported	Unknown
88	Grok 4 0709xAI	44.48%	Not reported	Unknown
89	Gemini 1.5 Flash 002Google	42.77%	Not reported	Unknown
90	Grok 4.1 Fast (Reasoning)xAI	42.61%	Not reported	Closed
91	Grok 4 Fast (Reasoning)xAI	42.09%	Not reported	Closed
92	Meta Llama Llama 3.2 90B Vision Instruct TurboTogether AI	38.79%	Not reported	Unknown
93	Mistral Medium 2505Mistral AI	36.45%	Not reported	Unknown
94	Grok 4.1 Fast Non ReasoningxAI	34.26%	Not reported	Unknown
95	Mistral Medium 3.5Mistral AIhigh reasoning	28.89%	Not reported	Unknown
96	Grok 4 Fast Non ReasoningxAI	25.32%	Not reported	Unknown
97	Meta Llama Llama 3.2 11B Vision Instruct TurboTogether AI	23.29%	Not reported	Unknown
98	Grok 2 Vision 1212xAI	3.46%	Not reported	Unknown

How MortgageTax is shown here

BenchLM mirrors the public Vals AI MortgageTax leaderboard captured from https://www.vals.ai/benchmarks/mortgage_tax and updated by Vals on September 1, 2026. The snapshot preserves overall scores, uncertainty, latency, cost-per-test metadata, and task-level scores where Vals publishes them.

MortgageTax 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

98 Vals rows3 task viewsprivate datasetTasks: Overall, Semantic Extraction, Numerical ExtractionDisplay only

MortgageTax on Vals AI Vals methodology Vals home

The published MortgageTax snapshot places Claude Opus 5 first at 72.06%. The third row is 1.79 points behind. The broader top-10 range is 3.30 points, so many of the published results sit in a relatively narrow band.

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

About MortgageTax

Year

2026

Tasks

Mortgage and tax extraction tasks

Format

Accuracy score

Difficulty

Professional mortgage-tax document reasoning

BenchLM mirrors Vals MortgageTax as a display-only finance and document-reasoning benchmark.

MortgageTax Public benchmark source

Freshness and provenance

Version

MortgageTax 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.

Questions

What does MortgageTax measure?

Vals AI benchmark for mortgage and tax document reasoning, including semantic and numerical extraction task views.

Which model leads the published MortgageTax snapshot?

Claude Opus 5 currently leads the published MortgageTax snapshot with 72.06% mortgagetax score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on MortgageTax?

The September 1, 2026 snapshot contains 98 AI models.

Last updated: September 1, 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 7,000+ subscribers.

One email each week. Unsubscribe anytime.