V5.3 // NEURAL_PIPELINE_ONLINE

STATE_OF_THE_ART AI_MEMORY ZERO_CLOUD_DEPENDENCY AIR_GAPPED

LongMemEval_Score Locally Hosted Leaderboard

LongMemEval_Score Overall Leaderboard

Air-gapped memory retrieval. Thousands of sessions. Zero cloud dependency. Echo_Locate runs a full index-filter-rank-synthesize pipeline on consumer hardware, matching cloud-scale systems on standardized benchmarks while keeping every byte local.

Request_Pilot ›› Tech_White_Paper

LONGMEMEVAL_BENCHMARK_ANALYSIS

OFFICIALLY_GPT-4o_JUDGED

01 · MASTRA_OM · GPT-5-MINI CLOUD — SOTA_CLOUD 94.87

02 · MASTRA_OM · GEMINI-3-PRO CLOUD 93.27

03 · HINDSIGHT · GEMINI-3-PRO CLOUD 91.40

04 · MASTRA_OM · GEMINI-3-FLASH CLOUD 89.20

05 · HINDSIGHT · GPT-OSS-120B CLOUD 89.00

06 · ECHO_LOCATE · QWEN_3.6_27B LOCAL_SOTA 85.40

07 · SUPERMEMORY · GEMINI-3-PRO CLOUD 85.20

08 · SUPERMEMORY · GPT-5 CLOUD 84.60

09 · MASTRA_OM · GPT-4o CLOUD 84.23

10 · HINDSIGHT · GPT-OSS-20B CLOUD 83.60

11 · EMERGENCEMEM_SIMPLE · GPT-4o CLOUD 82.40

12 · ECHO_LOCATE · QWEN_3.5_27B LOCAL 82.40

13 · ZEP_GRAPHITI · GPT-4o CLOUD 71.20

OVERALL_RANK#06

QA_SCORE85.40 %

LOCALLY_HOSTED / AIR_GAPPED#01 OVERALL

READER_MODELQWEN3.6_27B_Q4_K_XL

COMPUTELOCAL

CLOUD_DEP.0.0 %

SYSTEM_DIAGNOSTICS

LIVE · NODE-01

CORE_TEMP78 °C

MODEL_SIZE27B

GPU3090 — 38 TOK/S

MODEL_CLASS✓ OPEN_WEIGHT

LONGMEMEVAL_BENCHMARK_ANALYSIS

QUANTIZED_27B_MODELS_ON_DECADE-OLD_EQUIPMENT_BREAKING_INTO_THE_LEADERBOARD
AGAINST_FRONTIER_1T+_MODELS_HOSTED_IN_CLOUD_DATA_CENTERS

OFFICIALLY_GPT-4o_JUDGED · SINGLE_PASS_QA · VERIFIED_LEADERBOARD · ALL_NUMBERS_CROSS-CHECKED

#	SYSTEM / RUN	READER_MODEL	COMPUTE	QA_SCORE · 4o_JUDGED	NOTES
01	MASTRA_OM	GPT-5-MINI	CLOUD_SOTA	94.87%	HIGHEST_EVER_SOTA
02	MASTRA_OM	GEMINI-3-PRO-PREVIEW	CLOUD	93.27%	—
03	HINDSIGHT	GEMINI-3-PRO-PREVIEW	CLOUD	91.40%	—
04	MASTRA_OM	GEMINI-3-FLASH-PREVIEW	CLOUD	89.20%	—
05	HINDSIGHT	GPT-OSS-120B	CLOUD	89.00%	—
06	ECHO_LOCATE	QWEN_3.6_27B	LOCAL	85.40%	LOCAL · AIR_GAPPED · SOTA
07	SUPERMEMORY	GEMINI-3-PRO-PREVIEW	CLOUD	85.20%	—
08	SUPERMEMORY	GPT-5	CLOUD	84.60%	—
09	MASTRA_OM	GPT-4o	CLOUD	84.23%	—
10	HINDSIGHT	GPT-OSS-20B	CLOUD	83.60%	—
11	EMERGENCEMEM_SIMPLE	GPT-4o	CLOUD	82.40%	—
11	ECHO_LOCATE	QWEN_3.5_27B	LOCAL	82.40%	LOCAL · AIR_GAPPED
13	SUPERMEMORY	GPT-4o	CLOUD	81.60%	—
14	MASTRA_RAG	GPT-4o	CLOUD	80.05%	—
15	ZEP_GRAPHITI	GPT-4o	CLOUD	71.20%	—
16	FULL_CONTEXT	GPT-4o	CLOUD	60.20%	CLOUD_BASELINE

■ LOCAL / AIR_GAPPED ■ CLOUD / 3P_HOSTED 16_VERIFIED_ENTRIES · GPT-4o_JUDGED · SINGLE_PASS_QA

FOOTNOTE — EmergenceMem Internal (86.00%) is intentionally excluded from this leaderboard. Mastra's research page documents that this configuration is “not publicly reproducible.” We exclude unreproducible results so every entry on this board can be independently verified. EmergenceMem's reproducible Simple configuration is included at 82.40%. SRC: mastra.ai/research/observational-memory · supermemory.ai/research

PER_CATEGORY_HEAD_TO_HEAD

CATEGORY_BY_CATEGORY_BREAKDOWN_AGAINST_THE_TOP_CLOUD_HOSTED_SYSTEMS

8_SYSTEMS · 6_CATEGORIES · SORTED_HIGH_TO_LOW_WITHIN_CATEGORY · GPT-4o_JUDGED

ECHO_LOCATE CLOUD_COMPETITORS

CAT_01

TEMPORAL_REASONING

01MASTRA_OM · GPT-5-MINI95.50%

02ECHO_LOCATE · QWEN_3.6_27B92.48%

03MASTRA_OM · GPT-4o85.70%

04SUPERMEMORY · GEMINI-3-PRO81.95%

05SUPERMEMORY · GPT-581.20%

06SUPERMEMORY · GPT-4o76.69%

07ZEP62.40%

08FULL_CONTEXT45.10%

CAT_02

KNOWLEDGE_UPDATE

01MASTRA_OM · GPT-5-MINI96.20%

02ECHO_LOCATE · QWEN_3.6_27B89.74%

02SUPERMEMORY · GEMINI-3-PRO89.74%

04SUPERMEMORY · GPT-4o88.46%

05SUPERMEMORY · GPT-587.18%

06MASTRA_OM · GPT-4o85.90%

07ZEP83.30%

08FULL_CONTEXT78.20%

CAT_03

MULTI_SESSION

01MASTRA_OM · GPT-5-MINI87.20%

02MASTRA_OM · GPT-4o79.70%

03SUPERMEMORY · GEMINI-3-PRO76.69%

04ECHO_LOCATE · QWEN_3.6_27B75.94%

05SUPERMEMORY · GPT-575.19%

06SUPERMEMORY · GPT-4o71.43%

07ZEP57.90%

08FULL_CONTEXT44.30%

CAT_04

SINGLE_SESSION_USER

01MASTRA_OM · GPT-4o98.60%

02SUPERMEMORY · GEMINI-3-PRO98.57%

03SUPERMEMORY · GPT-597.14%

03SUPERMEMORY · GPT-4o97.14%

05ECHO_LOCATE · QWEN_3.6_27B95.71%

06MASTRA_OM · GPT-5-MINI95.70%

07ZEP92.90%

08FULL_CONTEXT81.40%

CAT_05

SINGLE_SESSION_ASSISTANT

01SUPERMEMORY · GPT-5100.00%

02SUPERMEMORY · GEMINI-3-PRO98.21%

03SUPERMEMORY · GPT-4o96.43%

04MASTRA_OM · GPT-5-MINI94.60%

04FULL_CONTEXT94.60%

06ECHO_LOCATE · QWEN_3.6_27B85.71%

07MASTRA_OM · GPT-4o82.10%

08ZEP80.40%

CAT_06

SINGLE_SESSION_PREFERENCE

01MASTRA_OM · GPT-5-MINI100.00%

02SUPERMEMORY · GPT-576.67%

03MASTRA_OM · GPT-4o73.30%

04SUPERMEMORY · GEMINI-3-PRO70.00%

04SUPERMEMORY · GPT-4o70.00%

06ECHO_LOCATE · QWEN_3.6_27B60.00%

07ZEP56.70%

08FULL_CONTEXT20.00%

FIG 1.0 PATH: /CORE/CORTEX/SCHEMA SYS: NEURAL_NETWORK_V5.3

Echo's_Neural_Pipeline

TOPOGRAPHIC MAP ● MODEL_SIZE: 27B ● MODEL_TYPE: OPEN_WEIGHT ● MODE: LOCAL_HOSTED

// GENERATION

QWEN3.6_27B – DUAL_INSTANCES

MODELQWEN3.6-27B

QUANTQ4_K_XL

SIZE_ON_DISK~17 GB

VRAM_LOADED~21.5 GB

CONTEXT_KV262K

HARDWARE: RTX_3090 x2

CHASSIS: HP_Z840

// EMBEDDING

QWEN3_EMBED_4B – SHARED_EMBEDDING

MODELQWEN3-EMBEDDING-4B

QUANTQ8_8BIT

SIZE_ON_DISK~4.1 GB

VRAM_LOADED~4.8 GB

RUNTIMELOCAL

HARDWARE: RTX_5060_TI

CHASSIS: LENOVO_THINKSTATION_S20

// RERANKER

QWEN3_RERANK_4B – SHARED_RERANKER

MODELQWEN3-RERANKER-4B

QUANTFP16_FULL

LOADED_VIAAUTOMODEL_FOR_CAUSALLM

SIZE_ON_DISK~8 GB

VRAM_LOADED~8.1 GB

HARDWARE: RTX_5060_TI

CHASSIS: LENOVO_THINKSTATION_S20

SECURE_DOMAINS

LEGAL_DISCOVERY

PRIVILEGED / E-DISC

Summarize, redact, and cite millions of privileged documents inside the firm’s enclave.

◆ CHAIN_OF_CUSTODY

HEALTHCARE_PHI

HIPAA / HITECH / EMA

PHI-safe diagnostics and molecular search running beside the MRI, never to a tenant.

◆ PHI_CONTAINED

DEFENSE_INTEL

DOD / ITAR / IL6

Classified inference for tactical edge, autonomy loops, and sensor fusion under air-gap.

◆ CLEARED_DEPLOY

QUANT_FINANCE

SEC / FINRA / SOC2-II

Low-latency signal engines and risk reasoners that never leave the trading enclave.

◆ AUDIT_READY

$2K_NODE_SPEC

CHASSIS_INFO

DEPLOYED_CHASSIS // FIELD_UNITS // SCRAP_WORKSTATIONS VS FRONTIER_CLOUD

HP Z840 workstation tower — HP_Z840 2x_RTX_3090 · QWEN3.6_27B_Q4_K_XL_DUAL_INSTANCES
2x_XEON_E5-2650_V4
64GB_DDR4_RAM
512GB_NVME_HARD_DRIVE
UBUNTU_OPERATING_SYSTEM

Lenovo ThinkStation S20 workstation tower — LENOVO_THINKSTATION_S20 RTX_5060_TI · EMBED_4B + RERANK_4B
1x_INTEL_XEON_W3680
16GB_DDR3_RAM
128GB_SATA_SSD_HARD_DRIVE
UBUNTU_OPERATING_SYSTEM

SYSTEM_LOG // node-01 TAIL -F

PILOT_ENROLLMENT_PROGRAM

INTERESTED IN LEARNING MORE? PLEASE SHARE YOUR USE CASE AND CONTACT INFO BELOW.

STATE_OF_THE_ART AI_MEMORY ZERO_CLOUD_DEPENDENCY AIR_GAPPED

LONGMEMEVAL_BENCHMARK_ANALYSIS

PER_CATEGORY_HEAD_TO_HEAD

TEMPORAL_REASONING

KNOWLEDGE_UPDATE

MULTI_SESSION

SINGLE_SESSION_USER

SINGLE_SESSION_ASSISTANT

SINGLE_SESSION_PREFERENCE

Echo's_Neural_Pipeline

QWEN3.6_27B – DUAL_INSTANCES

QWEN3_EMBED_4B – SHARED_EMBEDDING

QWEN3_RERANK_4B – SHARED_RERANKER

SECURE_DOMAINS

LEGAL_DISCOVERY

HEALTHCARE_PHI

DEFENSE_INTEL

QUANT_FINANCE

$2K_NODE_SPEC

PERFORMANCE

POWER_USAGE

PILOT_ENROLLMENT_PROGRAM