Skip to main content
Radar

Five or fewer confirmed AI changes, with original sources, on mornings when something changed.A free source-linked morning brief.

Start free brief
BenchLM
Model A
Agents-A1

InternScience

61.3/100

Estimated · Public rank #53

90% interval 51.4–71.1

Agents-A1 vs Kimi K2.6

Updated August 21, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.

Model B
Kimi K2.6

Moonshot AI

60.2/100

Estimated · Public rank #61

90% interval 50.4–70.1

Decision reading

Agents-A1 has the higher public score estimate, 61.25 versus 60.24, but the 90% score intervals overlap. Treat that as a lead, not a settled winner.

2 results are shared. Category rows based on different benchmark sets are marked directional and do not name a winner.

Which one for your work

Recommendations appear only when a shared evidence basis or an explicit operating constraint supports the call. Secondary and unsupported use cases stay disclosed below the initial list.

  • Long documents

    Prompts that approach the documented context limit

    Agents-A1

    Agents-A1 has the larger documented context window.

    Confidence: documented

Show secondary and unsupported calls

  • Coding work

    Code generation, repair, and software-engineering tasks

    Not enough matched evidence

    No shared weighted benchmark basis supports a winner.

    Confidence: limited

  • Agentic work

    Tool use, computer use, and multi-step task completion

    Not enough matched evidence

    The category averages use different weighted benchmark sets, so they are directional rather than like-for-like.

    Confidence: limited

  • Chat turn cost

    1K fresh input + 500 output tokens

    Not enough matched evidence

    A complete comparable API-rate estimate is not available for both models.

    Confidence: listed-rates

  • Cache-heavy agent loop cost

    200K cached + 20K fresh input + 10K output tokens

    Not enough matched evidence

    A complete comparable API-rate estimate is not available for both models.

    Confidence: rate-fallback

  • Repository review cost

    50K fresh input + 3K output tokens

    Not enough matched evidence

    A complete comparable API-rate estimate is not available for both models.

    Confidence: listed-rates

What is actually comparable

Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.

Shared results
2
Agents-A1 only
4
Kimi K2.6 only
30
Like-for-like categories
0 / 8

2 categories use different evidence sets. Those rows remain visible for coverage context but do not name a winner.

Category results, on a stated basis

Each row states whether both averages use the same weighted benchmark set. Directional and not-comparable rows remain visible, but they never receive a winner in this template.

Agentic

Directional only
Agents-A1
75.5
Kimi K2.6
73.5
Weighted basis
1 vs 3 rows
Reading
Directional only

Knowledge

Directional only
Agents-A1
47.6
Kimi K2.6
42.2
Weighted basis
1 vs 2 rows
Reading
Directional only

Coding

Not comparable
Agents-A1
Not measured
Kimi K2.6
64.4
Weighted basis
0 vs 3 rows
Reading
Not comparable

Reasoning

Not comparable
Agents-A1
60.2
Kimi K2.6
Not measured
Weighted basis
1 vs 0 rows
Reading
Not comparable

Math

Not comparable
Agents-A1
Not measured
Kimi K2.6
67.1
Weighted basis
0 vs 4 rows
Reading
Not comparable

Multilingual

Not comparable
Agents-A1
Not measured
Kimi K2.6
Not measured
Weighted basis
0 vs 0 rows
Reading
Not comparable

Multimodal

Not comparable
Agents-A1
Not measured
Kimi K2.6
79.8
Weighted basis
0 vs 2 rows
Reading
Not comparable

Instruction following

Not comparable
Agents-A1
94.8
Kimi K2.6
Not measured
Weighted basis
1 vs 0 rows
Reading
Not comparable

Shape of the matched evidence

Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.

Too few matched category axes support a radar. The ruled list below shows only shared benchmark results; positions use each benchmark’s normalized display scale when available.

What each workload costs

Three fixed token mixes turn per-token rates into comparable decisions. Each scenario states context fit and whether cached input had to fall back to the published list-input rate.

Chat turn

1K fresh input + 500 output tokens

Agents-A1
API rate not published
Fits in one request
Kimi K2.6
$0.00295
Fits in one request

Agents-A1 has no comparable published API token rate.

Repository review

50K fresh input + 3K output tokens

Agents-A1
API rate not published
Fits in one request
Kimi K2.6
$0.0595
Fits in one request

Agents-A1 has no comparable published API token rate.

Cache-heavy agent loop

200K cached + 20K fresh input + 10K output tokens

Agents-A1
API rate not published
Fits in one request
Cached-input rate unavailable
Kimi K2.6
$0.249
Fits in one request
Cached input priced at the published list-input rate

Kimi K2.6 has no published cached-input rate, so cached tokens use its listed input rate. Agents-A1 has no comparable published API token rate.

Specification differences

Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.

Context window

Maximum documented context; output-token limits may be lower.

Agents-A1

262K

Kimi K2.6

256K

API model ID

Agents-A1

Not sourced

Kimi K2.6

Not sourced

Cached-input rate

A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.

Agents-A1

No comparable hosted API rate

Kimi K2.6

Not published

Documented inputs

Agents-A1

Not sourced

Kimi K2.6

Not sourced

Documented outputs

Agents-A1

Not sourced

Kimi K2.6

Not sourced

Provider availability

Agents-A1

Not sourced

Kimi K2.6

Not sourced

Reasoning profile

Agents-A1

Reasoning

Kimi K2.6

Reasoning

Weight access

Agents-A1

Open Weight

Kimi K2.6

Open Weight

License

Agents-A1

Open Weight

Kimi K2.6

Open Weight

Release date

Agents-A1

2026-06-26

Kimi K2.6

2026-04-20

If you already use one of these models
Deployment change
The models list different providers, so authentication, endpoint behavior, limits, and feature support may change.
Quality signal
Agents-A1 has the higher public score estimate, 61.25 versus 60.24, but the 90% score intervals overlap.
Workload cost
A complete comparable API-rate estimate is not available for both models.
Context tradeoff
Agents-A1 has the larger documented window (262K).

Run the same representative tasks against both endpoints before changing production traffic.

Self-host vs API cost

Estimates at 50,000 req/day · 1000 tokens/req average.

Agents-A1
API / mo$0
Self-host / moNot listed
Break-even—
Proprietary model — self-hosting not applicable.
Kimi K2.6
API / mo$3,713
Self-host / mo$18,221
Break-even326M/day
Model the full break-even

Benchmark evidence

The full public result ledger is available for audit without forcing a wide desktop table onto a phone.

Browse raw public benchmark evidence36 rows

Agentic

  • BrowseComp

    Agents-A175.5%
    Source
    Kimi K2.683.2%
    Source

    Kimi K2.6 leads this result

  • HLE w/ tools

    Agents-A147.6%
    Source
    Kimi K2.6—

    Not directly comparable

  • VITA-Bench

    Agents-A138.8%
    Source
    Kimi K2.6—

    Not directly comparable

  • Terminal-Bench 2.0

    Agents-A1—
    Kimi K2.666.7%
    Source

    Not directly comparable

  • OSWorld-Verified

    Agents-A1—
    Kimi K2.673.1%
    Source

    Not directly comparable

  • Toolathlon

    Agents-A1—
    Kimi K2.650%
    Source

    Not directly comparable

  • MCP Atlas

    Agents-A1—
    Kimi K2.655.9%
    Source

    Not directly comparable

  • Claw-Eval

    Agents-A1—
    Kimi K2.662.3%
    Source

    Not directly comparable

  • DeepSearchQA

    Agents-A1—
    Kimi K2.692.5%
    Source

    Not directly comparable

  • WideResearch

    Agents-A1—
    Kimi K2.680.8%
    Source

    Not directly comparable

  • Gert Labs

    Agents-A1—
    Kimi K2.656.82%
    Source

    Not directly comparable

  • ResearchClawBench

    Agents-A1—
    Kimi K2.618.0%
    Source

    Not directly comparable

  • OSWorld 2.0

    Agents-A1—
    Kimi K2.64.6%
    Source

    Not directly comparable

Coding

  • SWE-bench Verified

    Agents-A1—
    Kimi K2.680.2%
    Source

    Not directly comparable

  • LiveCodeBench v6

    Agents-A1—
    Kimi K2.689.6%
    Source

    Not directly comparable

  • SWE-bench Pro

    Agents-A1—
    Kimi K2.658.6%
    Source

    Not directly comparable

  • SWE Multilingual

    Agents-A1—
    Kimi K2.676.7%
    Source

    Not directly comparable

  • SciCode

    Agents-A1—
    Kimi K2.652.2%
    Source

    Not directly comparable

  • Terminal-Bench 2.0

    Agents-A1—
    Kimi K2.666.7%
    Source

    Not directly comparable

  • Vibe Code Bench

    Agents-A1—
    Kimi K2.637.89%
    Source

    Not directly comparable

  • cursorBench31

    Agents-A1—
    Kimi K2.647.6%
    Source

    Not directly comparable

Reasoning

  • LongBench v2

    Agents-A160.2%
    Source
    Kimi K2.6—

    Not directly comparable

Knowledge

  • HLE

    Agents-A147.6%
    Source
    Kimi K2.634.7%
    Source

    Agents-A1 leads this result

  • GPQA

    Agents-A1—
    Kimi K2.690.5%
    Source

    Not directly comparable

  • GPQA-D

    Agents-A1—
    Kimi K2.690.5%
    Source

    Not directly comparable

Math

  • AIME26

    Agents-A1—
    Kimi K2.696.4%
    Source

    Not directly comparable

  • HMMT Feb 2026

    Agents-A1—
    Kimi K2.692.7%
    Source

    Not directly comparable

  • MMAnswerBench

    Agents-A1—
    Kimi K2.686.0%
    Source

    Not directly comparable

  • FrontierMath v2 (Tiers 1-3)

    Agents-A1—
    Kimi K2.638.966%
    Source

    Not directly comparable

  • FrontierMath v2 (Tier 4)

    Agents-A1—
    Kimi K2.614.580%
    Source

    Not directly comparable

Multimodal

  • MMMU-Pro

    Agents-A1—
    Kimi K2.679.4%
    Source

    Not directly comparable

  • MMMU-Pro w/ Python

    Agents-A1—
    Kimi K2.680.1%
    Source

    Not directly comparable

  • CharXiv

    Agents-A1—
    Kimi K2.680.4%
    Source

    Not directly comparable

  • MathVision

    Agents-A1—
    Kimi K2.687.4%
    Source

    Not directly comparable

  • V*

    Agents-A1—
    Kimi K2.696.9%
    Source

    Not directly comparable

Instruction following

  • IFEval

    Agents-A194.8%
    Source
    Kimi K2.6—

    Not directly comparable

Frequently asked questions

Which is better, Agents-A1 or Kimi K2.6?

Agents-A1 has the higher public score estimate, 61.25 versus 60.24, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.

Which is better for coding, Agents-A1 or Kimi K2.6?

The published evidence does not provide a shared weighted coding basis for both models, so BenchLM does not name a coding winner.

Which is better for agentic tasks, Agents-A1 or Kimi K2.6?

The current agentic tasks averages use different weighted benchmark sets, so BenchLM does not name a winner from them. Read the shared benchmark rows directly and test the models on the same task set.

Which costs less, Agents-A1 or Kimi K2.6?

Both models do not have comparable published API token rates, so this page does not name a universal price winner.

Which has the larger context window, Agents-A1 or Kimi K2.6?

Agents-A1 has the larger documented context window: 262K, compared with 256K.

Related comparisons

Compare API pricing Read the methodology Open the model selector

Last updated August 21, 2026

Watch Agents-A1 vs Kimi K2.6

One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.

Read a sample issue

Join 2,000+ readers.