Vals Finance Agent v2

Agentic financial research evaluates whether a system can use tools and evidence to complete analyst work.27 Oct 2026

One row per model and compatible score measure. Mean = equal-weight arithmetic average of imported reports. Distinct configurations reported for the same model in a source table remain separate. Reports can overlap; the range is not a confidence interval. Unbundle to inspect each observation.
		Evidence	What the result shows
Gemini 4 Argon →Google	65.4% 2 reports	Vals Finance Agent v2[1][3] View 2 reports 65.4% · Vals Finance Agent v2 [1] 65.4 % · Finance Agent v2 · Accuracy · Cost / Test: $4.38; Input / Output Cost: $4 / $20; Duration: 16m21s [3]	The imported reports agree at the published precision. Repeated citations may describe the same underlying evaluation. Leads this displayed comparison.
Gemini 3.8 Flash →Google	61.44% 1 report	Vals Finance Agent v2[3]Cost / Test: $2.00; Input / Output Cost: $1.5 / $7.5; Duration: 3m21s	#2 of 83 in this displayed comparison; 3.96 pp below the leader. $2.00 / test3m 21s
Muse Spark Open →Meta	61.34% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.77; Input / Output Cost: $1.25 / $4.25; Duration: 24m13s	#3 of 83 in this displayed comparison; 4.06 pp below the leader. $0.77 / test24m 13s
Muse Spark 1.2 →Meta	60.6% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.77; Input / Output Cost: $1.25 / $4.25; Duration: 5m04s	#4 of 83 in this displayed comparison; 4.8 pp below the leader. $0.77 / test5m 04s
Muse Spark 1.3 →MetaCost / Test: $0.76; Input / Output Cost: $1.25 / $4.25; Duration: 3m19s	59.96% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.76; Input / Output Cost: $1.25 / $4.25; Duration: 3m19s	#5 of 83 in this displayed comparison; 5.44 pp below the leader. $0.76 / test3m 19s
Muse Wonder Full →Meta	59.1% 1 report	Vals Finance Agent v2[2]	#6 of 83 in this displayed comparison; 6.3 pp below the leader.
Wonder Full →	59.06% 1 report	Vals Finance Agent v2[3]Cost / Test: $2.39; Input / Output Cost: $4 / $20; Duration: 14m56s	#7 of 83 in this displayed comparison; 6.34 pp below the leader. $2.39 / test14m 56s
Gemini 3.7 Flash →Google	59.04% 1 report	Vals Finance Agent v2[3]Cost / Test: $1.48; Input / Output Cost: $1.5 / $7.5; Duration: 3m01s	#8 of 83 in this displayed comparison; 6.36 pp below the leader. $1.48 / test3m 01s
Muse Spark 1.3 →MetaCost / Test: $0.74; Input / Output Cost: $1.25 / $4.25; Duration: 5m42s	58.9% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.74; Input / Output Cost: $1.25 / $4.25; Duration: 5m42s	#9 of 83 in this displayed comparison; 6.5 pp below the leader. $0.74 / test5m 42s
Claude Fable 5.1 →Anthropic	58.89% 3 reports · range 58.88–58.9%	Vals Finance Agent v2[1][2][3] View 3 reports 58.9% · Vals Finance Agent v2 [1] 58.9 · Vals Finance Agent v2 [2] 58.88 % · Finance Agent v2 · Accuracy · Cost / Test: $8.35; Input / Output Cost: $10 / $50; Duration: 18m14s [3]	Published observations span 0.02 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty. #10 of 83 in this displayed comparison; 6.51 pp below the leader.
Claude Opus 5 →Anthropic	58.63% 1 report	Vals Finance Agent v2[3]Cost / Test: $5.12; Input / Output Cost: $5 / $25; Duration: 9m58s	#11 of 83 in this displayed comparison; 6.77 pp below the leader. $5.12 / test9m 58s
Claude Opus 5.5 →Anthropic	58.6% 3 reports · range 58.59–58.6%	Vals Finance Agent v2[1][2][3] View 3 reports 58.6% · Vals Finance Agent v2 [1] 58.6 · Vals Finance Agent v2 [2] 58.59 % · Finance Agent v2 · Accuracy · Cost / Test: $9.22; Input / Output Cost: $4 / $20; Duration: 33m08s [3]	Published observations span 0.01 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty. #12 of 83 in this displayed comparison; 6.8 pp below the leader.
Claude Sonnet 5.5 →Anthropic	58.1% 1 report	Vals Finance Agent v2[3]Cost / Test: $6.55; Input / Output Cost: $2 / $10; Duration: 31m43s	#13 of 83 in this displayed comparison; 7.3 pp below the leader. $6.55 / test31m 43s
Qwen 3.8 Max 0902 →Qwen	57.87% 1 report	Vals Finance Agent v2[3]Cost / Test: $3.03; Input / Output Cost: $2 / $6; Duration: 49m05s	#14 of 83 in this displayed comparison; 7.53 pp below the leader. $3.03 / test49m 05s
Gemini 3.5 Flash →Google	57.86% 1 report	Vals Finance Agent v2[3]Cost / Test: $2.51; Input / Output Cost: $1.5 / $9; Duration: 5m22s	#15 of 83 in this displayed comparison; 7.54 pp below the leader. $2.51 / test5m 22s
GLM 5.3 Flash →Z.ai	57.85% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.05; Input / Output Cost: $0.075 / $0.25; Duration: 14m51s	#16 of 83 in this displayed comparison; 7.55 pp below the leader. $0.05 / test14m 51s
MiMo V2.6 Pro →Xiaomi	57.34% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.20; Input / Output Cost: $0.435 / $0.87; Duration: 10m48s	#17 of 83 in this displayed comparison; 8.06 pp below the leader. $0.20 / test10m 48s
Muse Spark 1.1 →Meta	57.21% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.72; Input / Output Cost: $1.25 / $4.25; Duration: 6m55s	#18 of 83 in this displayed comparison; 8.19 pp below the leader. $0.72 / test6m 55s
DeepSeek V4.1 Flash 1017 →DeepSeek	56.96% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.46; Input / Output Cost: $0.3 / $1.2; Duration: 14m39s	#19 of 83 in this displayed comparison; 8.44 pp below the leader. $0.46 / test14m 39s
Wonder →	56.64% 1 report	Vals Finance Agent v2[3]Cost / Test: $2.67; Input / Output Cost: $4 / $20; Duration: 6m09s	#20 of 83 in this displayed comparison; 8.76 pp below the leader. $2.67 / test6m 09s
Muse Wonder →Meta	56.6% 1 report	Vals Finance Agent v2[2]	#21 of 83 in this displayed comparison; 8.8 pp below the leader.
Claude Fable 5 →Anthropic	56.31% 1 report	Vals Finance Agent v2[3]Cost / Test: $8.06; Input / Output Cost: $10 / $50; Duration: 10m11s	#22 of 83 in this displayed comparison; 9.09 pp below the leader. $8.06 / test10m 11s
Gemini 3.6 Flash →Google	56.3% 1 report	Vals Finance Agent v2[3]Cost / Test: $1.40; Input / Output Cost: $1.5 / $7.5; Duration: 3m45s	#23 of 83 in this displayed comparison; 9.1 pp below the leader. $1.40 / test3m 45s
MiMo V2.6 Flash →Xiaomi	56.28% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.07; Input / Output Cost: $0.14 / $0.28; Duration: 7m02s	#24 of 83 in this displayed comparison; 9.12 pp below the leader. $0.07 / test7m 02s
GLM 5.3 →Z.ai	55.84% 1 report	Vals Finance Agent v2[3]Cost / Test: $1.07; Input / Output Cost: $1.4 / $4.4; Duration: 15m50s	#25 of 83 in this displayed comparison; 9.56 pp below the leader. $1.07 / test15m 50s
Hy4 Preview →	55.06% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.59; Input / Output Cost: $0.834 / $2.501; Duration: 23m45s	#26 of 83 in this displayed comparison; 10.34 pp below the leader. $0.59 / test23m 45s
GPT-5.6 Luna →OpenAI	55.04% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.28; Input / Output Cost: $0.2 / $1.2; Duration: 12m52s	#27 of 83 in this displayed comparison; 10.36 pp below the leader. $0.28 / test12m 52s
Ling 3.0 Flash Fin →	54.93% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.04; Input / Output Cost: $0.06 / $0.18; Duration: 6m15s	#28 of 83 in this displayed comparison; 10.47 pp below the leader. $0.04 / test6m 15s
Mistral Large 4 →Mistral	54.68% 1 report	Vals Finance Agent v2[3]Cost / Test: $1.20; Input / Output Cost: $1.36 / $4.18; Duration: 26m02s	#29 of 83 in this displayed comparison; 10.72 pp below the leader. $1.20 / test26m 02s
GPT-5.6 Terra →OpenAI	54.44% 1 report	Vals Finance Agent v2[3]Cost / Test: $3.63; Input / Output Cost: $2 / $12; Duration: 24m10s	#30 of 83 in this displayed comparison; 10.96 pp below the leader. $3.63 / test24m 10s
Claude Haiku 5.5 →Anthropic	54.14% 1 report	Vals Finance Agent v2[3]Cost / Test: $1.64; Input / Output Cost: $0.1 / $0.5; Duration: 30m49s	#31 of 83 in this displayed comparison; 11.26 pp below the leader. $1.64 / test30m 49s
Claude Opus 4.8 →Anthropic	53.92% 1 report	Vals Finance Agent v2[3]Cost / Test: $4.22; Input / Output Cost: $5 / $25; Duration: 9m07s	#32 of 83 in this displayed comparison; 11.48 pp below the leader. $4.22 / test9m 07s
Claude Sonnet 5 →Anthropic	53.91% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.75; Input / Output Cost: $2 / $10; Duration: 13m12s	#33 of 83 in this displayed comparison; 11.49 pp below the leader. $0.75 / test13m 12s
GPT-5.6 Sol →OpenAI	53.76% 1 report	Vals Finance Agent v2[3]Cost / Test: $1.25; Input / Output Cost: $4 / $20; Duration: 19m25s	#34 of 83 in this displayed comparison; 11.64 pp below the leader. $1.25 / test19m 25s
Grok 4.6 →xAI	53.68% 1 report	Vals Finance Agent v2[3]Cost / Test: $1.66; Input / Output Cost: $2 / $6; Duration: 16m41s	#35 of 83 in this displayed comparison; 11.72 pp below the leader. $1.66 / test16m 41s
GPT-6 Astra →OpenAI	53.51% 3 reports · range 53.5–53.54%	Vals Finance Agent v2[3][1][2] View 3 reports 53.54 % · Finance Agent v2 · Accuracy · Cost / Test: $6.82; Input / Output Cost: $10 / $50; Duration: 11m33s [3] 53.5% · Vals Finance Agent v2 [1] 53.5 · Vals Finance Agent v2 [2]	Published observations span 0.04 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty. #36 of 83 in this displayed comparison; 11.89 pp below the leader.
DeepSeek V4.1 Flash →DeepSeek	53.48% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.21; Input / Output Cost: $0.3 / $1.2; Duration: 6m04s	#37 of 83 in this displayed comparison; 11.92 pp below the leader. $0.21 / test6m 04s
Kimi K3 →Kimi	53.11% 1 report	Vals Finance Agent v2[3]Cost / Test: $1.91; Input / Output Cost: $3 / $15; Duration: 4m46s	#38 of 83 in this displayed comparison; 12.29 pp below the leader. $1.91 / test4m 46s
Grok 4.7 →xAI	52.25% 1 report	Vals Finance Agent v2[3]Cost / Test: $2.72; Input / Output Cost: $2 / $6; Duration: 15m33s	#39 of 83 in this displayed comparison; 13.15 pp below the leader. $2.72 / test15m 33s
GPT-6.1 Sol →OpenAI	52.03% 1 report	Vals Finance Agent v2[3]Cost / Test: $1.62; Input / Output Cost: $2 / $10; Duration: 15m25s	#40 of 83 in this displayed comparison; 13.37 pp below the leader. $1.62 / test15m 25s
Ember-1 →	51.85% 1 report	Vals Finance Agent v2[3]Cost / Test: $1.95; Input / Output Cost: $3 / $15; Duration: 13m19s	#41 of 83 in this displayed comparison; 13.55 pp below the leader. $1.95 / test13m 19s
GPT-5.5 →OpenAI	51.76% 1 report	Vals Finance Agent v2[3]Cost / Test: $4.15; Input / Output Cost: $5 / $30; Duration: 11m02s	#42 of 83 in this displayed comparison; 13.64 pp below the leader. $4.15 / test11m 02s
Claude Opus 4.7 →Anthropic	51.51% 1 report	Vals Finance Agent v2[3]Cost / Test: $4.03; Input / Output Cost: $5 / $25; Duration: 6m00s	#43 of 83 in this displayed comparison; 13.89 pp below the leader. $4.03 / test6m 00s
Claude Sonnet 4.6 →Anthropic	51.03% 1 report	Vals Finance Agent v2[3]Cost / Test: $2.41; Input / Output Cost: $3 / $15; Duration: 11m36s	#44 of 83 in this displayed comparison; 14.37 pp below the leader. $2.41 / test11m 36s
Step 5 Preview →StepFun	50.67% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.62; Input / Output Cost: $1 / $2.7; Duration: 12m28s	#45 of 83 in this displayed comparison; 14.73 pp below the leader. $0.62 / test12m 28s
Qwen 3.8 Max →Qwen	50.59% 1 report	Vals Finance Agent v2[3]Cost / Test: $1.24; Input / Output Cost: $2 / $6; Duration: 19m35s	#46 of 83 in this displayed comparison; 14.81 pp below the leader. $1.24 / test19m 35s
DeepSeek V4 Pro 0813 →DeepSeek	50.39% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.88; Input / Output Cost: $1.32 / $3.96; Duration: 16m15s	#47 of 83 in this displayed comparison; 15.01 pp below the leader. $0.88 / test16m 15s
GPT-6 Luna →OpenAI	49.87% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.12; Input / Output Cost: $0.1 / $0.5; Duration: 14m58s	#48 of 83 in this displayed comparison; 15.53 pp below the leader. $0.12 / test14m 58s
GLM 5.2 →Z.ai	49.7% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.71; Input / Output Cost: $1.4 / $4.4; Duration: 7m53s	#49 of 83 in this displayed comparison; 15.7 pp below the leader. $0.71 / test7m 53s
DeepSeek V4 Flash 0731 →DeepSeek	49.52% 1 report	Vals Finance Agent v2[3]Cost / Test: $0.28; Input / Output Cost: $0.44 / $1.32; Duration: 8m10s	#50 of 83 in this displayed comparison; 15.88 pp below the leader. $0.28 / test8m 10s

Sources

Gemini 4 Argon Model Card ↗
Google DeepMind · 2026
Dates & evaluation details
Vals Finance Agent v2 · 2026-10-26
A transcription of the October model card, including its external leaderboard snapshots. These are not uniformly matched runs. Effort, tools, hardware, frame budgets and task subsets differ. GraphWalks reports F1; OSWorld uses the offline subset and partial credit. Harvey refers to the archived Vals v1 protocol, not the remediated v1.1 or AA variant. Consult the linked methodology.
Muse Wonder | Meta ↗
Meta AI
Dates & evaluation details
Vals Finance Agent v2 · 2026-10-26
Source-reported model settings are max unless a cell says otherwise. The original page reports numbers without a uniform percent symbol; original formatting is retained. OSWorld includes two scoring rules in each cell: leaderboard ordering uses partial credit only. The internal composite and AA index use separate scales; their scores cannot be averaged.
Finance Agent v2 — evaluation methodology and results ↗
Vals AI · 2026
Dates & evaluation details
Finance Agent v2 · Accuracy · 2026-10-26
Public source leaderboard captured 26 October 2026. Model names, configurations, native score units and source-reported uncertainty are retained. Version and task split follow this specific Vals methodology page. These are published observations, not evaluations run by RLTasks.
Introducing Kimi K3.1 Pro ↗
Moonshot AI · 2026
Dates & evaluation details
Vals Finance Agent v2 · Frontier Finance · Kimi K3.1 Pro · complete release comparison · 2026-10-27
Published model comparison, 2026. Kimi K3.1 Pro · complete release comparison. Exact values from the publisher’s embedded full comparison table. Maximum effort, fallback settings and dated model versions remain in model labels. Blank cells stay missing. Source footnotes specify mixed harnesses and externally quoted comparisons; this is a launch-reported cohort rather than a new independent evaluation. Frontier Finance uses the official finance-agent v2 harness. Benchmarks · full embedded comparison table · row 28; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

TRACE THE SCORE · Per-model notes use published breakdowns where available.

Score vs. cost per test

Vals AI · 77 observations

Finance Agent v2 · Accuracy · higher is better

Costs are Vals AI’s published Cost / Test in USD, paired with the score from the same result. They reflect that evaluation’s workload and configuration; they are not token prices or cost per successful task. Missing costs are omitted. Explore all reported results ↓

The dashed line traces the best observed score as reported cost increases. It describes this selection of reports, with their original configurations. It does not estimate unmeasured models or costs.

View plotted values and sources (77)

Vals Finance Agent v2 · Finance Agent v2 · Accuracy. Published measurements in their original score units.
Model	Score	Cost per test (USD)	Configuration & source
Gemini 4 Argon →	65.4 %	$4.38	Cost / Test: $4.38; Input / Output Cost: $4 / $20; Duration: 16m21s Vals AI · Finance Agent v2 · Accuracy ↗
Gemini 3.8 Flash →	61.44 %	$2.00	Cost / Test: $2.00; Input / Output Cost: $1.5 / $7.5; Duration: 3m21s Vals AI · Finance Agent v2 · Accuracy ↗
Muse Spark Open →	61.34 %	$0.77	Cost / Test: $0.77; Input / Output Cost: $1.25 / $4.25; Duration: 24m13s Vals AI · Finance Agent v2 · Accuracy ↗
Muse Spark 1.2 →	60.6 %	$0.77	Cost / Test: $0.77; Input / Output Cost: $1.25 / $4.25; Duration: 5m04s Vals AI · Finance Agent v2 · Accuracy ↗
Muse Spark 1.3 Max →	59.96 %	$0.76	Cost / Test: $0.76; Input / Output Cost: $1.25 / $4.25; Duration: 3m19s Vals AI · Finance Agent v2 · Accuracy ↗
Wonder Full →	59.06 %	$2.39	Cost / Test: $2.39; Input / Output Cost: $4 / $20; Duration: 14m56s Vals AI · Finance Agent v2 · Accuracy ↗
Gemini 3.7 Flash →	59.04 %	$1.48	Cost / Test: $1.48; Input / Output Cost: $1.5 / $7.5; Duration: 3m01s Vals AI · Finance Agent v2 · Accuracy ↗
Muse Spark 1.3 →	58.9 %	$0.74	Cost / Test: $0.74; Input / Output Cost: $1.25 / $4.25; Duration: 5m42s Vals AI · Finance Agent v2 · Accuracy ↗
Claude Fable 5.1 →	58.88 %	$8.35	Cost / Test: $8.35; Input / Output Cost: $10 / $50; Duration: 18m14s Vals AI · Finance Agent v2 · Accuracy ↗
Claude Opus 5 →	58.63 %	$5.12	Cost / Test: $5.12; Input / Output Cost: $5 / $25; Duration: 9m58s Vals AI · Finance Agent v2 · Accuracy ↗
Claude Opus 5.5 →	58.59 %	$9.22	Cost / Test: $9.22; Input / Output Cost: $4 / $20; Duration: 33m08s Vals AI · Finance Agent v2 · Accuracy ↗
Claude Sonnet 5.5 →	58.1 %	$6.55	Cost / Test: $6.55; Input / Output Cost: $2 / $10; Duration: 31m43s Vals AI · Finance Agent v2 · Accuracy ↗
Qwen 3.8 Max 0902 →	57.87 %	$3.03	Cost / Test: $3.03; Input / Output Cost: $2 / $6; Duration: 49m05s Vals AI · Finance Agent v2 · Accuracy ↗
Gemini 3.5 Flash →	57.86 %	$2.51	Cost / Test: $2.51; Input / Output Cost: $1.5 / $9; Duration: 5m22s Vals AI · Finance Agent v2 · Accuracy ↗
GLM 5.3 Flash →	57.85 %	$0.05	Cost / Test: $0.05; Input / Output Cost: $0.075 / $0.25; Duration: 14m51s Vals AI · Finance Agent v2 · Accuracy ↗
MiMo V2.6 Pro →	57.34 %	$0.20	Cost / Test: $0.20; Input / Output Cost: $0.435 / $0.87; Duration: 10m48s Vals AI · Finance Agent v2 · Accuracy ↗
Muse Spark 1.1 →	57.21 %	$0.72	Cost / Test: $0.72; Input / Output Cost: $1.25 / $4.25; Duration: 6m55s Vals AI · Finance Agent v2 · Accuracy ↗
DeepSeek V4.1 Flash 1017 →	56.96 %	$0.46	Cost / Test: $0.46; Input / Output Cost: $0.3 / $1.2; Duration: 14m39s Vals AI · Finance Agent v2 · Accuracy ↗
Wonder →	56.64 %	$2.67	Cost / Test: $2.67; Input / Output Cost: $4 / $20; Duration: 6m09s Vals AI · Finance Agent v2 · Accuracy ↗
Claude Fable 5 →	56.31 %	$8.06	Cost / Test: $8.06; Input / Output Cost: $10 / $50; Duration: 10m11s Vals AI · Finance Agent v2 · Accuracy ↗
Gemini 3.6 Flash →	56.3 %	$1.40	Cost / Test: $1.40; Input / Output Cost: $1.5 / $7.5; Duration: 3m45s Vals AI · Finance Agent v2 · Accuracy ↗
MiMo V2.6 Flash →	56.28 %	$0.07	Cost / Test: $0.07; Input / Output Cost: $0.14 / $0.28; Duration: 7m02s Vals AI · Finance Agent v2 · Accuracy ↗
GLM 5.3 →	55.84 %	$1.07	Cost / Test: $1.07; Input / Output Cost: $1.4 / $4.4; Duration: 15m50s Vals AI · Finance Agent v2 · Accuracy ↗
Hy4 Preview →	55.06 %	$0.59	Cost / Test: $0.59; Input / Output Cost: $0.834 / $2.501; Duration: 23m45s Vals AI · Finance Agent v2 · Accuracy ↗
GPT-5.6 Luna →	55.04 %	$0.28	Cost / Test: $0.28; Input / Output Cost: $0.2 / $1.2; Duration: 12m52s Vals AI · Finance Agent v2 · Accuracy ↗
Ling 3.0 Flash Fin →	54.93 %	$0.04	Cost / Test: $0.04; Input / Output Cost: $0.06 / $0.18; Duration: 6m15s Vals AI · Finance Agent v2 · Accuracy ↗
Mistral Large 4 →	54.68 %	$1.20	Cost / Test: $1.20; Input / Output Cost: $1.36 / $4.18; Duration: 26m02s Vals AI · Finance Agent v2 · Accuracy ↗
GPT-5.6 Terra →	54.44 %	$3.63	Cost / Test: $3.63; Input / Output Cost: $2 / $12; Duration: 24m10s Vals AI · Finance Agent v2 · Accuracy ↗
Claude Haiku 5.5 →	54.14 %	$1.64	Cost / Test: $1.64; Input / Output Cost: $0.1 / $0.5; Duration: 30m49s Vals AI · Finance Agent v2 · Accuracy ↗
Claude Opus 4.8 →	53.92 %	$4.22	Cost / Test: $4.22; Input / Output Cost: $5 / $25; Duration: 9m07s Vals AI · Finance Agent v2 · Accuracy ↗
Claude Sonnet 5 →	53.91 %	$0.75	Cost / Test: $0.75; Input / Output Cost: $2 / $10; Duration: 13m12s Vals AI · Finance Agent v2 · Accuracy ↗
GPT-5.6 Sol →	53.76 %	$1.25	Cost / Test: $1.25; Input / Output Cost: $4 / $20; Duration: 19m25s Vals AI · Finance Agent v2 · Accuracy ↗
Grok 4.6 →	53.68 %	$1.66	Cost / Test: $1.66; Input / Output Cost: $2 / $6; Duration: 16m41s Vals AI · Finance Agent v2 · Accuracy ↗
GPT-6 Astra →	53.54 %	$6.82	Cost / Test: $6.82; Input / Output Cost: $10 / $50; Duration: 11m33s Vals AI · Finance Agent v2 · Accuracy ↗
DeepSeek V4.1 Flash →	53.48 %	$0.21	Cost / Test: $0.21; Input / Output Cost: $0.3 / $1.2; Duration: 6m04s Vals AI · Finance Agent v2 · Accuracy ↗
Kimi K3 →	53.11 %	$1.91	Cost / Test: $1.91; Input / Output Cost: $3 / $15; Duration: 4m46s Vals AI · Finance Agent v2 · Accuracy ↗
Grok 4.7 →	52.25 %	$2.72	Cost / Test: $2.72; Input / Output Cost: $2 / $6; Duration: 15m33s Vals AI · Finance Agent v2 · Accuracy ↗
GPT-6.1 Sol →	52.03 %	$1.62	Cost / Test: $1.62; Input / Output Cost: $2 / $10; Duration: 15m25s Vals AI · Finance Agent v2 · Accuracy ↗
Ember-1 →	51.85 %	$1.95	Cost / Test: $1.95; Input / Output Cost: $3 / $15; Duration: 13m19s Vals AI · Finance Agent v2 · Accuracy ↗
GPT 5.5 →	51.76 %	$4.15	Cost / Test: $4.15; Input / Output Cost: $5 / $30; Duration: 11m02s Vals AI · Finance Agent v2 · Accuracy ↗
Claude Opus 4.7 →	51.51 %	$4.03	Cost / Test: $4.03; Input / Output Cost: $5 / $25; Duration: 6m00s Vals AI · Finance Agent v2 · Accuracy ↗
Claude Sonnet 4.6 →	51.03 %	$2.41	Cost / Test: $2.41; Input / Output Cost: $3 / $15; Duration: 11m36s Vals AI · Finance Agent v2 · Accuracy ↗
Step 5 Preview →	50.67 %	$0.62	Cost / Test: $0.62; Input / Output Cost: $1 / $2.7; Duration: 12m28s Vals AI · Finance Agent v2 · Accuracy ↗
Qwen 3.8 Max →	50.59 %	$1.24	Cost / Test: $1.24; Input / Output Cost: $2 / $6; Duration: 19m35s Vals AI · Finance Agent v2 · Accuracy ↗
DeepSeek V4 Pro 0813 →	50.39 %	$0.88	Cost / Test: $0.88; Input / Output Cost: $1.32 / $3.96; Duration: 16m15s Vals AI · Finance Agent v2 · Accuracy ↗
GPT-6 Luna →	49.87 %	$0.12	Cost / Test: $0.12; Input / Output Cost: $0.1 / $0.5; Duration: 14m58s Vals AI · Finance Agent v2 · Accuracy ↗
GLM 5.2 →	49.7 %	$0.71	Cost / Test: $0.71; Input / Output Cost: $1.4 / $4.4; Duration: 7m53s Vals AI · Finance Agent v2 · Accuracy ↗
DeepSeek V4 Flash 0731 →	49.52 %	$0.28	Cost / Test: $0.28; Input / Output Cost: $0.44 / $1.32; Duration: 8m10s Vals AI · Finance Agent v2 · Accuracy ↗
GPT-6 Sol →	49.05 %	$2.12	Cost / Test: $2.12; Input / Output Cost: $2 / $10; Duration: 9m00s Vals AI · Finance Agent v2 · Accuracy ↗
Qwen 3.8 27B →	48.55 %	$0.75	Cost / Test: $0.75; Input / Output Cost: $0.5 / $3; Duration: 10m22s Vals AI · Finance Agent v2 · Accuracy ↗
Grok 4.5 →	48.35 %	$1.16	Cost / Test: $1.16; Input / Output Cost: $2 / $6; Duration: 4m49s Vals AI · Finance Agent v2 · Accuracy ↗
MiniMax-M3 →	48.27 %	$0.32	Cost / Test: $0.32; Input / Output Cost: $0.6 / $2.4; Duration: 8m17s Vals AI · Finance Agent v2 · Accuracy ↗
Qwen 3.7 Max →	47.78 %	$0.55	Cost / Test: $0.55; Input / Output Cost: $2.5 / $7.5; Duration: 10m35s Vals AI · Finance Agent v2 · Accuracy ↗
Gemini 3.5 Flash Lite →	47.44 %	$0.39	Cost / Test: $0.39; Input / Output Cost: $0.3 / $2.5; Duration: 2m23s Vals AI · Finance Agent v2 · Accuracy ↗
Inkling →	46.6 %	$1.00	Cost / Test: $1.00; Input / Output Cost: $1 / $4.05; Duration: 18m02s Vals AI · Finance Agent v2 · Accuracy ↗
GPT 5.4 Mini →	45.36 %	$1.20	Cost / Test: $1.20; Input / Output Cost: $0.75 / $4.5; Duration: 35m54s Vals AI · Finance Agent v2 · Accuracy ↗
Kimi K2.6 →	44.9 %	$1.14	Cost / Test: $1.14; Input / Output Cost: $0.95 / $4; Duration: 27m10s Vals AI · Finance Agent v2 · Accuracy ↗
GLM 5.1 →	44.79 %	$0.62	Cost / Test: $0.62; Input / Output Cost: $1 / $3.2; Duration: 15m24s Vals AI · Finance Agent v2 · Accuracy ↗
DeepSeek V4 →	44.08 %	$0.88	Cost / Test: $0.88; Input / Output Cost: $1.32 / $3.96; Duration: 25m26s Vals AI · Finance Agent v2 · Accuracy ↗
Gemini 3.1 Pro Preview (02/26) →	42.98 %	$1.59	Cost / Test: $1.59; Input / Output Cost: $2 / $12; Duration: 3m28s Vals AI · Finance Agent v2 · Accuracy ↗
Gemini 3 Flash (12/25) →	42.55 %	$0.57	Cost / Test: $0.57; Input / Output Cost: $0.5 / $3; Duration: 4m07s Vals AI · Finance Agent v2 · Accuracy ↗
MiMo V2.5 Pro →	41.5 %	$0.21	Cost / Test: $0.21; Input / Output Cost: $0.435 / $0.87; Duration: 8m17s Vals AI · Finance Agent v2 · Accuracy ↗
Inkling Small →	41.26 %	$0.23	Cost / Test: $0.23; Input / Output Cost: $0.3 / $1.2; Duration: 16m34s Vals AI · Finance Agent v2 · Accuracy ↗
Qwen 3.6 Plus →	40.85 %	$0.81	Cost / Test: $0.81; Input / Output Cost: $0.5 / $3; Duration: 10m12s Vals AI · Finance Agent v2 · Accuracy ↗
Qwen 3.7 Plus →	38.22 %	$0.36	Cost / Test: $0.36; Input / Output Cost: $0.4 / $1.6; Duration: 9m02s Vals AI · Finance Agent v2 · Accuracy ↗
GPT 5.4 Nano →	38.22 %	$0.16	Cost / Test: $0.16; Input / Output Cost: $0.2 / $1.25; Duration: 5m35s Vals AI · Finance Agent v2 · Accuracy ↗
Grok 4.3 →	37.73 %	$0.85	Cost / Test: $0.85; Input / Output Cost: $1.25 / $2.5; Duration: 10m15s Vals AI · Finance Agent v2 · Accuracy ↗
MiMo V2.5 →	36.73 %	$0.09	Cost / Test: $0.09; Input / Output Cost: $0.14 / $0.28; Duration: 5m39s Vals AI · Finance Agent v2 · Accuracy ↗
Kimi K2.5 →	35.79 %	$0.29	Cost / Test: $0.29; Input / Output Cost: $0.6 / $3; Duration: 12m32s Vals AI · Finance Agent v2 · Accuracy ↗
Mistral Medium 3.5 →	32.1 %	$1.91	Cost / Test: $1.91; Input / Output Cost: $1.5 / $7.5; Duration: 12m48s Vals AI · Finance Agent v2 · Accuracy ↗
Claude Haiku 4.5 (Thinking) →	31.01 %	$0.60	Cost / Test: $0.60; Input / Output Cost: $1 / $5; Duration: 2m57s Vals AI · Finance Agent v2 · Accuracy ↗
Ling 3.0 Flash →	30.31 %	$0.05	Cost / Test: $0.05; Input / Output Cost: $0.075 / $0.22; Duration: 2m37s Vals AI · Finance Agent v2 · Accuracy ↗
Gemini 3.1 Flash Lite Preview →	29.99 %	$0.14	Cost / Test: $0.14; Input / Output Cost: $0.25 / $1.5; Duration: 54.78s Vals AI · Finance Agent v2 · Accuracy ↗
Grok 4.20 (Reasoning) →	28.49 %	$0.79	Cost / Test: $0.79; Input / Output Cost: $2 / $6; Duration: 5m51s Vals AI · Finance Agent v2 · Accuracy ↗
MiniMax-M2.7 →	27.89 %	$0.18	Cost / Test: $0.18; Input / Output Cost: $0.3 / $1.2; Duration: 13m24s Vals AI · Finance Agent v2 · Accuracy ↗
Mercury 2.5 →	18.89 %	$0.11	Cost / Test: $0.11; Input / Output Cost: $0.2 / $0.75; Duration: 92.48s Vals AI · Finance Agent v2 · Accuracy ↗
Nemotron 3.5 Lightning →	18.5 %	$0.03	Cost / Test: $0.03; Input / Output Cost: $0.05 / $0.2; Duration: 9m00s Vals AI · Finance Agent v2 · Accuracy ↗

Reported in recent model launches

These model reports include this benchmark. Their harnesses, task subsets, scoring policies, and resource budgets can differ; follow each report’s evaluation notes before joining the numbers.

Kimi · Kimi K3.1 Pro →Meta · Muse Wonder and Muse Wonder Full →Google · Gemini 4 Argon →

What it measures

Agentic financial research evaluates whether a system can use tools and evidence to complete analyst work.

How to read the result

Check that the conclusion, calculations, and supporting documents agree, and preserve the final deliverable’s evaluation rubric.

Limits and comparison traps

A result belongs to the stated agent version and task distribution. Financial benchmark performance is not live investment performance.

Published evaluation

The linked publication documents the task and scoring procedure. Model results retain the source’s evaluation setting, native score units and available uncertainty.

Published evaluation

Exact values from the publisher’s embedded full comparison table. Maximum effort, fallback settings and dated model versions remain in model labels. Blank cells stay missing. Source footnotes specify mixed harnesses and externally quoted comparisons; this is a launch-reported cohort rather than a new independent evaluation. Frontier Finance uses the official finance-agent v2 harness.

Where this fits in an evaluation suite

Synthetic research rewards do not establish live investment performance. Preserve transaction costs, held-out regimes, constraints, and the exact policy behind a reported result.

Start with a capability hypothesis, then choose a second test with a different failure mode. Keep the checkpoint, reasoning effort, agent tools, sampling policy, and scoring procedure fixed when measuring a training intervention. Record failures and missing outputs in the denominator.

A reproducible comparison

Publish the exact run configuration, item coverage, score units, and available uncertainty. If a source supplies only an aggregate or a single run, preserve that limit. A new model name or a rounded score alone does not establish a capability improvement.

PRIMARY EVIDENCE & FURTHER READING

Introducing Kimi K3.1 Pro ↗
Moonshot AI · 2026
Finance Agent v2 — evaluation methodology and results ↗
Vals AI · 2026
Muse Wonder | Meta ↗
Meta AI

Reviewed 2026-10-27. Explanation and interpretation by RLTasks. Benchmark ownership remains with its original authors.

Continue exploring

All financial reasoning benchmarks →

Sources

Score vs. cost per test

Reported in recent model launches

What it measures

How to read the result

Limits and comparison traps

Published evaluation

Published evaluation

Where this fits in an evaluation suite

A reproducible comparison

Continue exploring

τ³-Banking

BankerToolBench

OfficeQA Pro

FinQA