GPT-5.4
Last updated: September 15, 2026
GPT-5.4 completed 29 of 55 private-markets finance tasks (52.7%) at $0.71 per task and 39 seconds each, ranking #15 of 17 models tested — strongest on metric lookups, weakest on document ingestion.
Tested on 55 tasks across seven kinds of financial work — reading documents, looking up figures, building statements, finding and fixing errors, assembling dashboards, drawing charts, and multi-step requests. It passed 11 of 12 on metric lookups and 1 of 6 on document ingestion. Run-to-run noise is about ±1.8 tasks, so gaps of two tasks or fewer are ties.
Overall Rank
#15
of 17 engines tested
Pass Rate
52.7%
29 of 55 tasks passed
Judge Score
31.6
out of 55.0 available
Suite Split
28/49
analysis suite · 1/6 ingest harness
Cost per Task
$0.71
8th cheapest of 17
Cost, Full Pass
$39
all 55 tasks at list prices
Cost per Passed Task
$1.35
9th best value of 17
Latency
39s
per task · 36 min for a full pass
Category Breakdown
GPT-5.4 against the average of every engine tested, category by category. The gap between the two bars is the only thing worth reading here — a category where the whole field struggles is a property of the tasks, not of the engine.
- Field average
- This model
| Category | Passed | Pass Rate | Field Avg | Rank | Score | Cost/Task | Latency | Tool Calls |
|---|---|---|---|---|---|---|---|---|
| Document Ingestion | 1/6 | 16.7% | 42.2% | #13 of 17 | 1.52 | $1.34 | 91s | 27.3 |
| Metric Lookups | 11/12 | 91.7% | 88.7% | #8 of 17 | 11.10 | $0.19 | 15s | 8.9 |
| Financial Statements | 2/7 | 28.6% | 43.7% | #13 of 17 | 2.77 | $0.40 | 31s | 17.4 |
| Error Detection & Repair | 2/7 | 28.6% | 60.5% | #15 of 17 | 2.15 | $0.68 | 54s | 25.0 |
| Dashboards & Widgets | 5/8 | 62.5% | 84.6% | #15 of 17 | 5.06 | $0.97 | 48s | 24.8 |
| Charts | 3/4 | 75.0% | 66.2% | #1 of 17 | 3.00 | $0.83 | 43s | 21.0 |
| Multi-Step Requests | 3/6 | 50.0% | 54.9% | #9 of 17 | 3.71 | $1.45 | 41s | 19.0 |
Every column is measured within that category. The ingest harness is a separate suite of longer, heavier workflows — several times the tool calls of an analysis task — so its cost and latency are not comparable with the six rows below it.
Where it fell short
The 23 tasks GPT-5.4 did not pass, of 50 graded individually. A benchmark that publishes only totals has to be taken on trust; these are the specific misses behind this model's score.
| Task | Kind of work | What it tests | Result |
|---|---|---|---|
| DD-01 | Document Ingestion | PDF → new rack, 5 components to the cent | Failed |
| DD-03 | Document Ingestion | Reuse an existing dimension, create no duplicates | Failed · 0.20 |
| TBU-01 | Document Ingestion | 556-account trial balance, signed, zero rows skipped | Failed |
| RCN-02 | Document Ingestion | P&L workbook, exclude subtotals and calculated rows | Failed |
| RCN-03 | Document Ingestion | 87 stores × 9 metrics, dimensioned, totals excluded | Failed · 0.32 |
| DIML-04 | Metric Lookups | — | Failed · 0.10 |
| CTA-03 | Financial Statements | — | Failed · 0.15 |
| CTA-04 | Financial Statements | — | Failed · 0.31 |
| CTA-05 | Financial Statements | — | Failed · 0.12 |
| SBM-01 | Financial Statements | — | Failed |
| TRACE-01 | Financial Statements | — | Failed · 0.23 |
| DEN-01 | Error Detection & Repair | — | Failed |
| DEN-02 | Error Detection & Repair | — | Failed |
| DISC-01 | Error Detection & Repair | — | Failed |
| DISC-02 | Error Detection & Repair | — | Failed · 0.15 |
| DISC-03 | Error Detection & Repair | — | Failed |
| BDATE-01 | Dashboards & Widgets | — | Failed |
| BOARD-01 | Dashboards & Widgets | — | Failed |
| WVB-01 | Dashboards & Widgets | — | Failed · 0.20 |
| CHART-02 | Charts | — | Failed · 0.03 |
| CMP-03 | Multi-Step Requests | — | Failed · 0.46 |
| MT-02 | Multi-Step Requests | — | Failed |
| MT-03 | Multi-Step Requests | — | Failed · 0.25 |
A fraction means the task was run more than once and the runs disagreed. A number after a failure is the partial credit awarded. Five of the 55 tasks in the headline total are not broken out individually in the published results.
Against the Field
Every engine on the same axis, with GPT-5.4 picked out. Grey bars are the rest of the field.
Overall Pass Rate
#15 of 17 · higher is better
Cost — One Full Pass (55 tasks)
#8 of 17 · lower is better
Cost per Passed Task
#9 of 17 · lower is better
Latency — Average per Task
#1 of 17 · lower is better
Tool Calls — Average per Task
#1 of 17 · lower is leaner
Cost vs Pass Rate
Up and to the left is better
Henon Tasks vs Tools benchmark — Ingest v5.1 + analysis suite v4, 17 executor models, 935 graded runs. Last updated September 15, 2026. Full results at henon.ai/benchmarks