In many long-horizon settings, an agent's own choices create future constraints: committing finite capacity establishes future obligations and forecloses later opportunities. Sustaining performance therefore requires both keeping evolving plans feasible and comparing the value of resource allocations with the alternatives they displace. Recent benchmarks increasingly evaluate LLM agents in continuing operations, but give less attention to how agents keep their own commitments feasible while reassessing resource allocations as opportunities change. We introduce OutfitterBench, where agents manage an outdoor-equipment rental business for 200 days, purchasing from a 46-product catalog and handling roughly 1,100 reservation requests. Under a fixed capital ceiling, reservations and uncertain walk-in demand compete for reusable equipment whose availability changes over time. Experiments with seven frontier LLMs show that several agents underperform a simple rule-based baseline, while only the leading agent consistently surpasses it. Trajectory analysis shows that some agents apply known constraints inconsistently and make new decisions that conflict with existing commitments. Agents with more reliable fulfillment can still secure low returns from comparable commitments or delay resource reallocation. The leading agent better connects preparation with economic choices but also misses opportunities for more profitable asset use. These findings highlight limitations in sustaining coherent resource plans and reassessing their value over time. The benchmark code is publicly available at https://github.com/adminblackjack/OutfitterBench.
Figures & tables
Figure 1: Business value created over the season, averaged equally over two world seeds for each agent. Curves use contract-adjusted economic value relative to the initial shop, accounting for equipment and outstanding obligations. Labels on the right report mean official net value creation ( ΔV ) at day 214. Shading marks the 14-day automatic settlement period. Calendar provides a fixed-policy reference between the highest- and lower-performing agents.
Figure 2: The OutfitterBench rental shop combines two demand channels with a reusable equipment fleet. The agent manages bookings, walk-in prices and occupancy limits, and equipment investment using business records, public forecasts, and customer feedback. Calendar and weather jointly shape demand and the timing of equipment reuse, while a fixed equipment-capital ceiling constrains fleet expansion. Performance after 200 decision days and 14 automatic settlement days is terminal cash plus eligible equipment resale value, less outstanding liabilities.
Figure 3: Competing uses of a single tent. The new request overlaps a confirmed booking, whereas a walk-in rental could fit before it with a one-day margin after cleaning. Late return or servicing delays could eliminate this margin. Solid and dashed outlines denote confirmed and candidate rentals, respectively.
Agent
ΔV (CNY)
Bookings
Reliab. (%)
Compens. (CNY)
GPT 6.1 Sol
+2,556.78
253.5
95.86
241.22
GPT 6 Luna
+947.44
206.5
93.75
228.68
MiMo V2.6 Flash
+736.29
263.0
84.55
802.64
Claude Opus 5.5
+636.11
296.5
90.70
472.71
Kimi K3
+154.55
238.5
80.68
987.00
GLM 5.2
-231.16
243.5
72.58
1,578.26
Table 1: Full-season outcomes averaged equally across the two world seeds. ΔV is value created above the initial business value; bookings are reservation pickups, including agreed reduced delivery. Reliability is the mean of the two run-level rates. Appendix C.1 reports the individual outcomes for each world.
Figure 4: Diagnostics for LLM agents, averaged equally across two seeds. A reports projected gaps in pending bookings; B measures new gaps introduced into existing bookings by subsequent actions and walk-in rentals. C measures net service contribution per occupied capital-time. D summarizes estimated gains from relaxing walk-in limits under a conservative capacity check with a three-day turnaround allowance. It reports the share of improving dates and the mean gain across all sampled dates, including zeros. These opportunities depend on each run’s fleet and limits. Section 4.4 defines the diagnostics; Appendix D gives calculation details and results by model and seed.
Appendix figures & tables11 assets
Supplementary material from the paper’s appendix.
Appendix
Component
Evaluated configuration
Decision / runoff days
200 / 14
Initial cash
3,000
Initial official value
3,397.06
Catalog / initial assets
46 SKUs / 27 items
Initial capital basis
1,141.83
Additional capital headroom
1,000.00
Appendix
Table 2: Common study configuration. Monetary amounts are simulated CNY; equipment-capital headroom is distinct from cash.
Agent
Model calls
Tool calls
Input tokens (M)
Output tokens (M)
Seed 42
Sol
2,524
5,835
305.26
2.58
Luna
3,279
8,850
251.81
13.63
MiMo
2,904
4,659
279.13
14.38
Opus
2,452
3,047
150.61
2.52
Kimi
2,251
3,390
138.34
4.83
Appendix
Table 3: Full-season model and tool calls, with input and output tokens in millions. A model response can request multiple tools, so the call counts differ.
Tool
Type
Purpose
Business records and public information
business_status
R
Inspect finances, obligations, assets, resource limits, and day progress.
business_history
R
Review completed-day business totals and successful actions.
finance_history
R
Review financial totals and daily financial records.
service_history
R
Review rental-service outcomes and product-level performance.
counter_log
R
Read recent individual walk-in interactions and their outcomes.
Appendix
Table 4: All 43 tools available to the evaluated agents. R: read-only; W: changes business or memory state; End: closes the business day. Descriptions summarize the evaluated operations.
Agent
ΔV
Bookings
Reliab. (%)
Compens.
Seed 42
Sol
+2,460.38
263
92.93
426.33
Luna
+1,237.24
188
92.16
282.12
MiMo
+1,198.11
266
81.85
992.47
Opus
+610.03
289
91.46
372.64
Kimi
+403.26
221
81.85
829.57
Appendix
Table 5: Full-season outcomes within each world. Monetary values are CNY; ΔV is the change from the initial business value of 3,397.06. Bookings are reservation pickups, including agreed reduced delivery; reliability is pickups divided by pickups plus merchant-caused failures.
Agent
Reservation rent
Walk-in rent
Damage deductions
Net equipment outlay
Operating expenses
Seed 42
Sol
7,344.36
2,609.75
218.98
1,223.21
6,092.44
Luna
5,326.23
2,510.27
193.49
1,024.41
5,371.28
MiMo
6,058.99
2,949.25
283.20
1,002.19
6,694.08
Opus
5,677.90
1,818.25
318.56
834.22
5,973.40
Kimi
5,682.52
2,340.00
237.78
1,283.25
6,176.73
Appendix
Table 6: Full-season income and expenditure in CNY. The three income columns, less net equipment outlay, operating expenses, and the common initial equipment value of 397.06, yield ΔV in Table 5 . Estimated usage wear is not an additional expense here.
Agent
All pending bookings
1–2 days before pickup
Seed 42
Sol
8.76 (198/2260)
13.70 (74/540)
Luna
9.17 (141/1537)
12.18 (47/386)
MiMo
26.52 (590/2225)
34.59 (211/610)
Opus
5.96 (153/2565)
8.83 (53/600)
Kimi
22.56 (446/1977)
30.24 (150/496)
Appendix
Table 7: Projected gaps: percentage (gapped/observed booking–days). A booking can contribute on multiple days. The near-pickup subset is determined at each observation, not a fixed recovery cohort.
Agent
Confirm
Quote
Fixed offer
Walk-in
Seed 42
Sol
3.13 (3/96)
3.33 (2/60)
6.98 (6/86)
0.50 (1/201)
Luna
7.50 (3/40)
8.16 (4/49)
2.63 (2/76)
2.23 (4/179)
MiMo
15.22 (14/92)
11.29 (7/62)
9.09 (8/88)
4.13 (10/242)
Opus
1.55 (2/129)
2.74 (2/73)
4.48 (3/67)
1.76 (3/170)
Kimi
23.68 (18/76)
14.58 (7/48)
10.98 (9/82)
3.28 (6/183)
Appendix
Table 8: Individual results behind Figure 4 B: percentage (gap-introducing/exposed units). Confirm, Quote, and Fixed offer use relevant successful calls; Walk-in uses actual customer rental batches. Each unit counts once even if several bookings acquire gaps. Types are not pooled. Definitions are in Appendix D.3 .
Agent
Net service contribution (CNY)
Occupied capital (CNY-days)
Productivity
Seed 42
Sol
5,917.79
430,562.99
1.374
Luna
4,687.77
448,281.77
1.046
MiMo
4,430.20
443,423.04
0.999
Opus
3,820.33
421,113.66
0.907
Kimi
4,048.36
419,729.18
0.965
Appendix
Table 9: Capital productivity: net service contribution per 100 capital-CNY-days. Contribution deducts estimated use consumption but excludes fixed operating fees and equipment purchase/sale cash flows. Capital-days include basis awaiting next-day release.
Agent
g=1 Mean gain
Days
g=3 Mean gain
Days
g=7 Mean gain
Days
Seed 42
Sol
1.210
9
0.654
5
0.064
1
Luna
2.190
13
1.556
10
0.667
4
MiMo
1.217
12
0.500
6
0.453
5
Opus
1.007
11
0.609
7
0.227
3
Kimi
1.727
14
0.929
9
0.623
5
Appendix
Table 10: Capacity-relaxation tests on 40 sampled days per run. For each turnaround allowance, Mean gain is estimated CNY improvement averaged over all sampled days, including zero and no-candidate days; Days counts those with an improvement of at least 0.01 CNY. The main figure uses g=3 .
Agent
Bookings checked
g=0 (%)
g=1 (%)
g=2 (%)
Seed 42
Sol
302
98.01
95.70
78.15
Luna
215
96.74
93.02
78.60
MiMo
339
94.99
84.96
67.26
Opus
337
99.41
98.81
72.11
Kimi
285
95.09
82.46
67.02
Appendix
Table 11: New-booking calendar coverage at day end under three turnaround allowances. The checked set is identical across allowances within each run; confirmations terminated that day are excluded.
Seed
Pairs
Opus rent
Sol rent
Opus below Sol (%)
42
44
823.90
1,183.50
30.38
73
33
480.05
763.17
37.10
Appendix
Table 12: Total agreed rent (CNY) for matched fully delivered bookings. Opus charges less in 75 of 77 pairs; the percentage difference uses Sol’s total rent as its base. Matching criteria are in Appendix E.2.1 .