Loading your app…

SWE-ContextBench

Make0 on SWE-ContextBench.

The comparison below is SWE-ContextBench Lite: 99 related issues against 300 earlier coding sessions. Make0 Standard’s overall score is 93.9%, 28.1% better than LangMem, the previous best at 73.3%. It returns the gold past task at rank 1 on 71.7% of related issues. This is Standard, the mode available to all users, not Max mode. LoCoMo results are on their own page.

Overall

0.0%

28.1% better than LangMem

Gold task in the top 10 · Lite · evaluated 25 Oct 2026

Rank 1

0.0%

Gold past task · Make0 Standard · Lite

Rank 1

Gold-task match at rank 1

Share of related issues whose first retrieved memory is the gold past task. Higher bars are better.

SWE-ContextBench top-1 gold-task match for Make0 Standard, LangMem, OpenViking, free summary search, Mem0, and Supermemory

Overall

Gold task anywhere in the results

Share of related issues where the gold past task appears anywhere in that system’s list. Each bar uses that system’s own result count. Higher bars are better.

SWE-ContextBench overall gold-task match for Make0 Standard, LangMem, OpenViking, free summary search, Mem0, and Supermemory

Comparison

Where the gold task lands

Each issue has one gold past task, so it can sit in only one slot. Rank 1 is the score that matters: higher is better. Rank 2 and rank 3 count the issues Make0 did not get first. A lower number there is the result of a strong rank 1, not a weaker system. A higher rank-2 or rank-3 number is not a better score.

SWE-ContextBench gold-task match by rank for Make0 Standard and published memory systems
MethodRank 1Gold at rank 2Gold at rank 3Overall
Make0 (Standard)Top 1071.7%14.1%2.0%93.9%
LangMemTop 1055.6%35.4%14.1%73.3%
OpenVikingTop 338.4%35.4%40.4%51.5%
Free summary searchAgent chooses34.3%2.0%0.0%36.4%
Mem0Top 325.3%15.2%9.1%39.4%
SupermemoryTop 1523.2%20.2%11.1%59.6%

The table is Make0 Standard, not Max mode, on SWE-ContextBench Lite (300 experience tasks, 99 related tasks), evaluated 25 Oct 2026. Every row in the table is Lite. Overall is the share of issues where the gold past task appears in that system’s retrieved results.

Rank 2 and rank 3 count the issues Make0 did not get first. A lower number there is the result of a strong rank 1, not a weaker system. A higher rank-2 or rank-3 number is not a better score.

LangMem, OpenViking, Mem0, Supermemory, and free summary search are Table 5 of SWE-ContextBench. Overall uses each system’s own result count. Make0 and LangMem are both over 10 results, so LangMem’s 73.3% is the same depth as Make0. Supermemory’s 59.6% is over 15. This table is retrieval of the gold past task on Lite. It is not the paper’s bug-resolution rate.

Full SWE-ContextBench

357 related issues · 1007 past tasks. Published trajectories cover 300 of those past tasks; the other memories are the issue text. Evaluated 25 Oct 2026.

Overall

0.0%

Rank 1

0.0%

Start with one Brain

Free to start

Build your Brain in two minutes

Sign up, add a few memories, and connect your first agent. That's the whole setup.

Make0Your BrainClaudeChatGPTCursorCursorGrok BotMake0Make0 Agent