Navigator Bench · Index v2.0 · September 2026 (demo)

Demo data · not real results

Top-rank quality at 1/4 the cost per successful task

5 models share rank 1 (none is significantly better than another on the same tasks); Model F · low effort (Index 93.6) costs $0.028 per successful task, Model D · high effort (Index 94.5) $0.13.

Navigator Index (0–100) — higher is better
Cost per successful task (log scale) — lower is better
Navigator Bench · Index v2.0 · data through 28 Sep 2026Up is better quality, left is cheaper, colour is provider.Demo data

Data through 28 Sep 20261,560 runsIndex v2.0ChangelogMethodChart card 16:91:1

02Leaderboard

Pick a category. Watch the order change.

Top-5 card

Ordered by rank. Ranks are ranges: 2–4 means we can't statistically separate it from the models in that range (paired test on the same tasks; the chance of any false win on the board is at most 5%). The bar under each rank shows where its range sits.

Model
Categories
Actions
1–5
Model DProvider 2 · high effort
94.593.0–95.9
$0.13
26 s
=+0.9
1–5
Model DProvider 2 · low effort
93.692.1–95.1
$0.050
11 s
=+0.4
1–5
Model FProvider 3 · low effort
93.691.8–95.3
$0.028
7.6 s
=+0.5
1–5
Model AProvider 1 · low effort
93.091.1–94.8
$0.057
8.8 s
=−0.7
1–5
Model IProvider 4 · low effort
93.090.9–95.1
$0.032
9.8 s
=±0.0
6–8
Model KProvider 5 · low effort
79.876.2–83.6
$0.0056
7.4 s
=+1.6
6–8
Model BProvider 1 · low effort
79.476.2–82.8
$0.016
5.5 s
=+6.5
6–8
Model GProvider 3 · low effort
73.869.2–78.1
$0.0082
3.1 s
=−4.3
9–13
Model CProvider 1 · low effort
50.743.5–57.4
$0.010
2.6 s
=+5.0
9–13
Model LProvider 5 · low effort
49.043.3–54.9
$0.0062
3.6 s
=−2.2
9–13
Model EProvider 2 · low effort
47.341.3–53.1
$0.0083
2.2 s
=+3.0
9–13
Model JNEWProvider 4
46.540.8–52.4
$0.014
2.4 s
NEW
9–13
Model HProvider 3
42.837.8–48.5
$0.0064
1.5 s
▲1+1.4

Navigator Bench · Index v2.0 · data through 28 Sep 2026Whiskers: 95% task-clustered intervals · How ties work · Cost per success

03Kill zone

Better and cheaper, or it doesn't count

Model K · low effort dominates 5 of 12 models on Navigator Bench — higher Index, lower cost per successful task.

Showing the largest kill zone this cycle. Pick any model here, from a point's card, or with the ⊘ button in the leaderboard; the scatter and the table dim with it.

Download card

A kill needs both: significantly better on the Index (a paired test on the same tasks, with the chance of any false win on the board held to 5%) and cheaper per successful task. A gap inside the noise never counts. How dominance is decided

  1. 1–5Model D · high94.5 · $0.13
  2. 1–5Model D · low93.6 · $0.050
  3. 1–5Model F · low93.6 · $0.028
  4. 1–5Model A · low93.0 · $0.057
  5. 1–5Model I · low93.0 · $0.032
  6. 6–8Model K · low79.8 · $0.0056
  7. 6–8Model B · low79.4 · $0.016
  8. 6–8Model G · low73.8 · $0.0082
  9. 9–13Model C · low50.7 · $0.010dominated
  10. 9–13Model L · low49.0 · $0.0062dominated
  11. 9–13Model E · low47.3 · $0.0083dominated
  12. 9–13Model J46.5 · $0.014dominated
  13. 9–13Model H42.8 · $0.0064dominated

04Commoditization curve

What Index 70 costs, cycle by cycle

Curve card

Cheapest at Index ≥ 70 · September 2026 (demo)

$0.0056

per successful task · Model K · low effort

▼ 2% since August 2026 (demo)

Tracking since 28 Aug 2026. The curve grows one point per published cycle; earlier history is never back-filled.

Cheapest cost per successful task at Index ≥ 70 (log scale) — lower is better
Tracking since 28 Aug 2026 · 2 published cycles so far
Published cycle (data cut-off)
Navigator Bench · Index v2.0 · data through 28 Sep 2026Each point: the cheapest model with Index ≥ 70 that cycleDemo data

September 2026 (demo): Model K · low effort was the cheapest model at Index 70 or better, at $0.0056 per successful task. Data through 28 Sep 2026.

05Compare

Two or three models, every category

CodingData analysisExtractionReasoningResearch & agenticSummarizationSupportWriting255075100
  • Model D · high effort
  • Model K · low effort
  • Shaded rings: 95% intervals
Model D · high effort
94.5Index · 93.0–95.9 · rank 1–5
Cost / success
$0.13
Median latency
26 s
Pass rate
100% (91%–100%)
Passes every time
100%
Model K · low effort
79.8Index · 76.2–83.6 · rank 6–8
Cost / success
$0.0056
Median latency
7.4 s
Pass rate
75% (64%–84%)
Passes every time
55%
Category · weightModel DModel K
Coding 15%92.586.0–98.1 · 5 tasks75.864.7–87.9 · 5 tasks
Data analysis 15%95.893.0–98.6 · 5 tasks68.356.2–81.4 · 5 tasks
Extraction 10%92.586.0–98.1 · 5 tasks84.273.9–92.6 · 5 tasks
Reasoning 15%98.396.5–100.0 · 5 tasks80.879.0–82.7 · 5 tasks
Research & agentic 15%For now these are single-call multi-step tasks: one model call returns plan → steps → deliverable, checked step by step. Tool-using, browsing and long agentic tasks are deferred.95.893.0–98.6 · 5 tasks86.778.3–95.1 · 5 tasks
Summarization 10%90.085.3–93.7 · 5 tasks75.059.2–90.8 · 5 tasks
Support 10%92.188.8–94.9 · 5 tasks80.870.6–90.2 · 5 tasks
Writing 10%96.792.9–100.0 · 5 tasks90.081.6–98.4 · 5 tasks
Run receipts120 demo runs · no receipts120 demo runs · no receipts

Per-task receipts (prompt, output, checks and cost for every run) open from each model's receipts link.

06Fit-finder

The cheapest model that clears your bar

Successful tasks a month

Cheapest at Navigator Index ≥ 70

Model K · low effort

$55.76a month · $0.0056 per successful task

Cheapest model at Navigator Index 70.0 or better: Model K · low effort, $0.0056 per successful task (about $55.76 for 10,000 successful tasks a month).

  1. Model K · lowNavigator Index score 79.8monthly cost $55.76/mo, clears the bar
  2. Model G · lowNavigator Index score 73.8~monthly cost $81.59/mo, clears the bar
  3. Model B · lowNavigator Index score 79.4monthly cost $165/mo, clears the bar
  4. Model F · lowNavigator Index score 93.6monthly cost $277/mo, clears the bar
  5. Model I · lowNavigator Index score 93.0monthly cost $324/mo, clears the bar
  6. Model D · lowNavigator Index score 93.6monthly cost $496/mo, clears the bar
  7. Model A · lowNavigator Index score 93.0monthly cost $569/mo, clears the bar
  8. Model D · highNavigator Index score 94.5monthly cost $1,313/mo, clears the bar
  9. Model L · lowNavigator Index score 49.0monthly cost $61.90/mo, below the bar
  10. Model HNavigator Index score 42.8monthly cost $63.51/mo, below the bar
  11. Model E · lowNavigator Index score 47.3monthly cost $83.16/mo, below the bar
  12. Model C · lowNavigator Index score 50.7monthly cost $101/mo, below the bar
  13. Model JNavigator Index score 46.5monthly cost $142/mo, below the bar

Bars are the Navigator Index score on a plain 0–100 scale; the tick is your bar, and the number beside each bar is the score itself. Monthly cost = cost per successful task × successful tasks, printed as a number. Failed runs are already priced in: cost per success divides everything spent by the tasks that passed. Definition

07Methodology & freshness

Every number, defined

28 Sep 2026data through
1,560runs
13entries
40tasks
v2.0Index
2 Oct 2026published

Demo data · not real results This board is built from development fixtures, not from a published cycle. Model names are placeholders and no number on it is a measurement.

Navigator Index
0–100. The weighted mean of 8 category scores; a category score is the mean task score (partial credit, 0–1) × 100 over that category’s active tasks and every repeat. Weights: Coding 15% · Data analysis 15% · Extraction 10% · Reasoning 15% · Research & agentic 15% · Summarization 10% · Support 10% · Writing 10%. Research & agentic: For now these are single-call multi-step tasks: one model call returns plan → steps → deliverable, checked step by step. Tool-using, browsing and long agentic tasks are deferred.
Intervals
95%, task-clustered: a bootstrap over tasks (stratified by category for the Index), never naive per-run intervals. Every score on this page shows its interval as a whisker or a range.
Ranks are ranges
Every pair of models is compared on the same tasks (a paired test of the Index difference, with family-wise error control, so the chance of any false “win” on the board is at most 5%). A rank is the range the data allows: best = 1 + models significantly better, worst = N − models significantly worse. “2–4” means we can't statistically separate it from the models in that range; a single number means it is pinned down.
Dominance (kill zone)
A model dominates another only when it is significantly better on the Index (the same paired test) and has a lower cost per successful task. A point-estimate gap inside the noise is never a kill. Reference rows are never counted.
Cost per successful task
Everything measured on candidate calls (OpenRouter's reported amounts) divided by the tasks passed, where a pass is a score at or above the task's threshold. Judge cost is excluded. Blank when nothing passed.
Latency
Median wall-clock latency per run, as measured on the pinned provider.
Movement
Δ compares with the previous published cycle at the same Index version only. Without one, the entry shows NEW or “first cycle”; movement is never estimated.
Commoditization curve
One point per published cycle: the cheapest cost per successful task among models at Index 70 or better. With fewer than three cycles it reads “tracking since”.
Frontier and quadrants
The frontier joins models that no other model beats on both Index and cost per success. Quadrants split at the quality bar (Index 70) and at the median plotted cost per success; they are reading aids, not verdicts.
Full methodologyChangelogReleasesCorrectionsIndependenceWhat changed in September 2026 (demo)Published comparisons (classic view)Presenter mode
DownloadsBoard API · available once a cycle is publishedJSON · available once a cycle is publishedCSV · available once a cycle is published