Results, Costs, and Marathons | Paper Mario Bench

Nine gates from save file to Bowser.

Each scored gate is one ninth of route progress. G1–G3 cover the opening; six castle steps lead to G4, when the first Bowser battle begins. Times below belong to the lifelong-player reference; S1u is an estimate.

Save

Save created

38.1s

Town

Toad Town

83.4s

Castle

Castle entry

110.1s

Hall

Castle hall

117.2s

Upper hall

Castle upper hall

~123.4s

Stairs

Castle stairs

129.6s

Upper

Upper castle

149.6s

Approach

Bowser approach

194.1s

Bowser

Bowser battle

227.4s

Compare the time spent inside the castle

The castle eats the clock

The opening gap is real. The castle gap is larger.

The lifelong-player human reference crossed G3 to G4 in 117.3 seconds. The fastest selected completion with exact castle timing needed 742.1 seconds.

Human1:57.3

GPT-6 Astra12:22.1

6.327× human castle time

Finishing hides the climb. Gate depth shows it.

Nine scored steps

Inspect a model, effort by effort.

Arithmetic mean of each selected attempt's deepest scorer gate on the nine-step G1–G4 route.

Progress model

Sort models by

G1SaveG2TownG3CastleS1HallS1uUpper hallS2StairsS3Upper castleS4ApproachG4Bowser

GPT-6 Astra
September 202615 selected · 15 G4
furthest G4avg 9.0/9
castle depth · furthest run
S1S1uS2S3S4G4
lown=3
3/3 G4G4 furthest9.0/9 avg
mediumn=3
3/3 G4G4 furthest9.0/9 avg
highn=3
3/3 G4G4 furthest9.0/9 avg
xhighn=3
3/3 G4G4 furthest9.0/9 avg
maxn=3
3/3 G4G4 furthest9.0/9 avg

Completion rates and gate timing by condition

Use this reference to compare effort cells within one model. Different harness conditions remain separate.

Completion by model × effort cell

Choose one model. Every bar keeps its effort and condition attached; mixed conditions are never pooled.

Model

low

G3 3/3median 25:48

0/3G4 / selected

medium

Not tested

high

G3 3/3median 21:00

0/3G4 / selected

xhigh

Not tested

max

G3 3/3median 23:23

0/3G4 / selected

G4 is not the whole story. GPT-5.6 Terra reaches the castle in 15 of 15 runs with a 12:20 median, so castle arrival is a given for it, yet only 5 of 15 finish. Gemini 3.6 Flash (8/9 to G3, median 10:20) reaches the castle consistently and never records a G4.

The selected field

247 attempts. 119 reached Bowser.

Each mark is one attempt. Choose a model, then an effort, to inspect its runs and finish times.

Which runs are counted?

The retained record contains 259 runs; 119 reached G4. The comparison includes 247 attempts, with 119 finishes; image-cap-blocked Muse Spark attempts are excluded. August and September 2026 remain separate collection periods, with harness and service differences retained.

The broader coverage map records 324 attempts across 24 model families and 85 cells. Its 19 marathon and side-experiment mappings stay separate from the comparison.

low: 85
medium: 64
high: 90
xhigh: 48
max: 37

Reached G4 No G4Showing 247 of 247 attempts.

Fastest selected20:05.3

GPT-6 Astra medium

Median completed47:32.5

12.544× human

Slowest completed89:47.2

Gemini 3.8 Flash medium

One dial, three stories

Effort is a setting, not a universal ladder.

The same low-to-max dial produced three different GPT-5.6 stories. Small cells, so treat them as stories, not proof.

Luna

4/15G4

high2/3reached G4

Sol

15/15G4

high3/3reached G4

Terra

5/15G4

high1/3reached G4

Three shapes, one family

In the August runs, these Anthropic models ran the same dial and drew three different curves. Each card plots average route progress at each effort setting.

Opus 5

The flat line

Every effort averages 100%. The dial moves the clock, not the depth.

Opus 4.8

The wobble

Medium dips, high recovers, xhigh dips again. More effort is not more depth.

Fable 5

The fade

Perfect at low and medium; high drops to 74% and xhigh stays there.

Muse Spark: testing blocked by the image cap

We tried Muse Spark 1.1, 1.2, and 1.3, including Contributor and non-Contributor variants. The API rejects requests with more than 50 images. That limit applies across Muse Spark, not just 1.1, and prevented comparable testing with this screenshot-driven harness. Muse Spark is therefore omitted from comparisons throughout the site.

The retained 1.1 attempts still show some of what happened before that limit: three runs ended mid-run because of it. 11 of 12 runs reached the castle, and the deepest made it two subgates past the door. Those observations remain available here, but they do not establish how well Muse Spark would perform without the image cap.

Separately, Muse Spark 1.1 kept trying to work outside the game, so I moved its runs into their own container. That containment history is described on the harness page.

Ninety minutes, sliced open

The game clock and wall clock tell different stories.

Frame time advances only while the game runs; the scored wall clock never stops. The harness page explains the pause machinery. The retained clocks cannot separate thinking from tooling from waiting, but they show how little of some runs was active game time.

Retained game-frame time Wall time outside the frame clock

Not inference latency. Time outside the naive frame clock mixes several systems and pause semantics. The scorer wall clock remains authoritative. The Kimi K3 example is a launch-window run that has since been re-run; it stays because no run rode the 90-minute wall closer.

The endurance field

8 marathon records. G4 was only the start.

The September six-hour runs add Astra and Fable 5.1 to the six earlier endurance records. Astra reached the first staircase in Koopa Bros. Fortress; Fable 5.1 reached the playground exit. Each bar follows ordered story checkpoints toward the first Star Spirit, with wall-clock time underneath. These checkpoints differ in difficulty and length. The segmented Fable 5 recovery lineage remains separate from single-session runs, and collection months stay visible.

Adjudicated valid; recovered disconnect · 1 segment

GPT-6 Astra medium

6:00:00.1

Reached G4: 25m
Deepest story state: Lowered first fortress staircase
No new depth at end: 3m
End: Time cap
Recorded usage estimate: ~$193

Story checkpoints 40 of 49The first fortress staircase is lowered. Checkpoints differ in difficulty and length; this is not a percentage of the game completed.

Retained token usage at Standard API rates below 272K context. Codex could record missing cache-write counts as zero; the range allows zero to all non-read input to incur the write premium. It excludes unreported final tails and orchestration overhead, and is not an invoice.

Reached the fortress and lowered its first staircase. A recovered transport warning was adjudicated valid; raw-invalid record and blocked filesystem-attempt caveat are preserved. Six-hour budget set before the run; Chapter 1 incomplete. Partial usage; final unreported requests are not treated as free.

Mario and Goombario stand behind a wall in the fortress prison corridor. — **Astra · Koopa Bros. Fortress**Astra ended in the fortress prison corridor. Earlier in the run, Astra lowered the first staircase.View original capture

Mario is in an enemy encounter on Goomba Road, with stars spinning around him. — **Astra · Koopa Bros. Fortress**Astra ended in the fortress prison corridor. Earlier in the run, Astra lowered the first staircase.View original capture

Almost tied at the opening. Hours apart after it.

Astra reached G4 at 25:04; Fable 5.1 at 24:35. Both had six hours at medium effort. By the end, Astra had lowered the first staircase inside Koopa Bros. Fortress. Fable's last new story milestone was leaving the playground, followed by continued combat and recovery attempts on Goomba Road.

Fable 5.1 wasn't the strongest Claude marathon result. An earlier six-hour Opus 5 run at medium effort reached the Toad Town story milestone, farther than Fable 5.1 managed here. Its trace from the August comparison provides a reference for the two September runs. The run details retain the differences in harness versions and automatic continuations.

Astra · SeptemberFable 5.1 · SeptemberOpus 5 · August

External scorer observations, shown as steps. One six-hour, medium-effort run per model, collected under different harness versions. Story checkpoints differ in length and difficulty. A flat line means no new recorded story state, not inactivity; a reload can move the line backward. None reached the first Star Spirit.

Read the milestone times and run caveats

Astra · medium · September 2026 comparison

Opening: 0:03:24
Wake in Goomba Village: 0:39:48
Found the hammer: 1:00:18
Defeated Jr. Troopa: 1:07:50
Left the playground: 1:09:31
Goombario joined: 1:25:59
Defeated Goomba Bros.: 2:04:02
Defeated Goomba King: 2:20:55
Arrived at Toad Town: 2:39:52
Met the Star Spirits: 3:02:31
Received the Lucky Star: 3:17:27
Reached Koopa Village: 4:18:25
Kooper joined: 4:40:19
Reached Koopa Bros. Fortress: 5:32:28
Defeated the basement guard: 5:54:43
Lowered the first staircase: 5:56:32

Raw-invalid transport disconnect was adjudicated valid after recovery; 530 successful commands followed.
Blocked prohibited-filesystem probe remains disclosed.
No observed subagent calls or controller memory-write commands in retained stream; this is not an exhaustive audit of arbitrary code paths.

Fable 5.1 · medium · September 2026 comparison

Opening: 0:02:35
Wake in Goomba Village: 0:53:19
Found the hammer: 2:33:36
Defeated Jr. Troopa: 2:41:21
Left the playground: 2:58:44

No recorded quota holds, scorer reattachments or mechanical continuations.
Blocked prohibited-filesystem probe remains disclosed.
No Task/Workflow subagent calls observed; all 393 accounted native receipts are primary, with zero sidechain receipts.
No controller memory-write commands observed; not an exhaustive audit of arbitrary code paths.

Opus 5 · medium · August 2026 comparison

Opening: 0:01:58
Wake in Goomba Village: 0:52:28
Found the hammer: 1:39:03
Defeated Jr. Troopa: 1:47:23
Left the playground: 1:52:02
Goombario joined: 2:35:05
Defeated Goomba Bros.: 3:52:58
Defeated Goomba King: 4:34:19
Arrived at Toad Town: 5:08:56

Historical reference from the August comparison; the raw run began July 25, 2026. Its harness version differs from the September runs; no clock normalization is applied.
One continuous external-scorer clock and one recorded game session, with two mechanical CLI continuations; no quota holds or scorer reattachments.
The curve uses retained external-scorer story observations, excluding three transient zero readings. Story milestones are ordinal, not percentages or equal amounts of work.
Deepest retained story milestone was arrival at Toad Town at 18,536 seconds; no later story milestone was recorded before the time cap.

The cost of keeping them playing.

For the two September runs, the retained usage gives API-equivalent estimates of ~$193 for Astra and $68.05 for Fable 5.1. Cache reads account for $168.51 of Astra's estimate. Fable's cheaper cache reads make a visible difference, although cache writes also contribute substantially to its total.

InputCache writesCache readsOutput

Astra~$193

Fable 5.1$68.05

Astra’s estimate treats all non-cached input as cache writes. Estimates use recorded tokens at the dated Standard API rates. Unreported final tails and orchestration overhead are excluded. These are neither subscription bills nor equal-progress costs: Astra got much farther.

Read the cost breakdown

Astra: estimated USD by token category
Category	Estimate
Input	$0.00
Cache writes	$16.34
Cache reads	$168.51
Output	$8.25
Total	$193.09

Fable 5.1: estimated USD by token category
Category	Estimate
Input	$0.12
Cache writes	$33.35
Cache reads	$22.53
Output	$12.05
Total	$68.05

Astra’s input/write split is estimated: all non-cached input is priced as cache writes, leaving $0 in ordinary input. Its actual write count was not reliably reported. This uses the upper end of the $189.82–$193.09 estimate range.

Fable 5.1: Completed native receipts with model-specific prices and cache TTL, including recorded helper models. Excludes unfinished or unobserved tails and aggregate-only usage. Not subscription spend or an invoice.

The score stops at G4.

The endurance story starts there.

Next: how the harness works

Nine gates from save file to Bowser.

Save

Town

Castle

Hall

Upper hall

Stairs

Upper

Approach

Bowser

The opening gap is real. The castle gap is larger.

Finishing hides the climb. Gate depth shows it.

Inspect a model, effort by effort.

247 attempts. 119 reached Bowser.

Effort is a setting, not a universal ladder.

The game clock and wall clock tell different stories.

Human control

Fable 5 low

GPT-5.5 xhigh

Kimi K3 Droid high

8 marathon records. G4 was only the start.

Almost tied at the opening. Hours apart after it.

Astra · medium · September 2026 comparison

Fable 5.1 · medium · September 2026 comparison

Opus 5 · medium · August 2026 comparison

The cost of keeping them playing.