T
THINKING MACHINES LAB / thinkingmachines/TML-d24-varloop-research

TML-d24-varloop-research

Research checkpoint released alongside Inkling 2 Mini on November 20. It allows the number of passes through reused middle layers to vary at inference. It retains 26.4 billion physical parameters and accepts text and images; it is not simply a quantization of Mini.

01 / SUMMARY

The essential

ORIENTATIONVariable-loop research

Primary use stated or inferred cautiously from official documentation.

CONTEXT1M tokens

Maximum input capacity when published by the source.

EXITText; separate output maximum unpublished

Documented maximum generation limit.

ACCESSHugging Face · downloadable checkpoint

Verified availability channels.

MODEL TYPEMultimodal · Code · Text and reasoning

Primary functional family and verified specialties.

LICENSEApache 2.0

Declared terms for API access, model weights, or self-hosting.

02 / TECHNICAL SHEET

Limits and integration

API IDthinkingmachines/TML-d24-varloop-research
Model typeMultimodal · Code · Text and reasoning
Access modelOpen weights
LicenseApache 2.0
DeploymentSelf-hosted
Release20 NOV 2026
Knowledge cutoffNot published
EntranceText · Image
ExitText
Context window1M tokens
maximum outputText; separate output maximum unpublished
ReasoningNot published
Published tools
Structured OutingsNot published
Batch processingNot published
Prompt cacheNot published
Fine-tuningNot published
Verified platformsHugging Face
BEST SUITED FOR

Study how quality changes when the same weights receive one, two, three, four or five passes.

WORTH MONITORING

Distributed as weights only: not available through an API or Tinker. Each additional pass adds computation, not new physical parameters. Use at least 32 GB of memory and compatible runtimes; do not attribute Mini’s scores to this checkpoint.

03 / CAPABILITIES

What it can do

01

Orientation

Study how quality changes when the same weights receive one, two, three, four or five passes.

02

Context

1M tokens · Text; separate output maximum unpublished

03

Tools and integration

Not published

04

Access

Hugging Face · downloadable checkpoint

Input modalities
TextImage
04 / PRICES

Documented cost

ConceptWorthUnit/condition
Standard inputNo published official API rateStandard API rate
Cached inputNot publishedReading reused prefixes
Cache write or storageNot publishedThe condition varies by provider
Standard outputSelf-hosted infrastructure costMay include reasoning tokens
Batch inputNot publishedAsynchronous processing
Batch outputNot publishedAsynchronous processing

No verified official API tariff. Self-hosting requires memory, compute and maintenance; downloadable weights do not make inference free.

i Prices change and may depend on level, region or context length. Check the source before making a decision.

05 / EVALUATIONS

How to read the results

“

At three passes, the official table reports 47.4% on DeepSWE v1.1 and 84.2% on GPQA Diamond. The pass count is retained because results change with it. Maxima from different configurations are not combined.

BenchmarkResultMetricSource
DeepSWE v1.1 (varloop 3 passes)47.4 %mini-swe-agent v2.4.6View source ↗
GPQA Diamond (varloop 3 passes)84.2 %Published scoreView source ↗

Inferama only highlights a “best result” when the metric, test set, configuration, and date allow for an equivalent comparison. The supplier's figures are presented as claims from its own source.

06 / VERSIONS

Chronology

thinkingmachines/TML-d24-varloop-research

Version added to Inferama's verified catalog.

07 / SOURCES

Traceability