Language modelOpen weights
K2 Horizon 7B
Institute of Foundation Models
- Released
- September 3, 2026
- Data date
- October 7, 2026
- Input
- Text
- Output
- Text
K2 Horizon 7B is a dense model with a 512K context for coding, reasoning, and multi-step tasks. After reinforcement learning, it receives an extra FTPO stage that the model card says is intended to reduce repetitive loops in long generations. The smaller 3.7B model follows a similar training path.
Specifications and access
| Specification | Value and source |
|---|---|
| Parameters | 7B total parametersSource |
| Active parameters | 7B parameters per tokenSource |
| Context window | 524,288 tokensSource |
| Input modalities | TextSource |
| Output modalities | TextSource |
| Documented training stages | Pretraining, four midtraining stages, math, code, search, and tool-use RL experts with a merge, two SFT phases, and a final anti-doom-loop FTPO step.Source |
| License | Apache 2.0Source |
Page 1 of 3
Measurements without matching peer values
These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.
AIME 2025 (Show measurement, test conditions, and source)
- Source value
- 90.3
- Score
- 90.3
- Metric
- AIME 2025
- Unit
- %
- Category
- source-specific
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- IFM model card
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Exact released K2 Horizon 7B anti_doom_loop_verified checkpoint after FTPO. IFM states before/after checkpoints use identical settings. The card supplies test-specific details: GPQA top_p=1 and 5 repeats, SciCode 3 repeats, SWE-bench via mini-swe-agent, Terminal-Bench via terminus-2 with 3 repeats at 256K context, BFCL temperature 0.01.
- Context
- Where the card gives a per-benchmark protocol, it is included in the configuration. Other rows do not have a reported sample count in the table.
AIME 2026 (Show measurement, test conditions, and source)
- Source value
- 90.2
- Score
- 90.2
- Metric
- AIME 2026
- Unit
- %
- Category
- source-specific
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- IFM model card
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Exact released K2 Horizon 7B anti_doom_loop_verified checkpoint after FTPO. IFM states before/after checkpoints use identical settings. The card supplies test-specific details: GPQA top_p=1 and 5 repeats, SciCode 3 repeats, SWE-bench via mini-swe-agent, Terminal-Bench via terminus-2 with 3 repeats at 256K context, BFCL temperature 0.01.
- Context
- Where the card gives a per-benchmark protocol, it is included in the configuration. Other rows do not have a reported sample count in the table.
HMMT February 2025 (Show measurement, test conditions, and source)
- Source value
- 87.7
- Score
- 87.7
- Metric
- HMMT February 2025
- Unit
- %
- Category
- source-specific
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- IFM model card
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Exact released K2 Horizon 7B anti_doom_loop_verified checkpoint after FTPO. IFM states before/after checkpoints use identical settings. The card supplies test-specific details: GPQA top_p=1 and 5 repeats, SciCode 3 repeats, SWE-bench via mini-swe-agent, Terminal-Bench via terminus-2 with 3 repeats at 256K context, BFCL temperature 0.01.
- Context
- Where the card gives a per-benchmark protocol, it is included in the configuration. Other rows do not have a reported sample count in the table.
HMMT February 2026 (Show measurement, test conditions, and source)
- Source value
- 77.8
- Score
- 77.8
- Metric
- HMMT February 2026
- Unit
- %
- Category
- source-specific
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- IFM model card
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Exact released K2 Horizon 7B anti_doom_loop_verified checkpoint after FTPO. IFM states before/after checkpoints use identical settings. The card supplies test-specific details: GPQA top_p=1 and 5 repeats, SciCode 3 repeats, SWE-bench via mini-swe-agent, Terminal-Bench via terminus-2 with 3 repeats at 256K context, BFCL temperature 0.01.
- Context
- Where the card gives a per-benchmark protocol, it is included in the configuration. Other rows do not have a reported sample count in the table.
GPQA Diamond (Show measurement, test conditions, and source)
- Source value
- 75.6
- Score
- 75.6
- Metric
- GPQA Diamond
- Unit
- %
- Category
- source-specific
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- IFM model card
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Exact released K2 Horizon 7B anti_doom_loop_verified checkpoint after FTPO. IFM states before/after checkpoints use identical settings. The card supplies test-specific details: GPQA top_p=1 and 5 repeats, SciCode 3 repeats, SWE-bench via mini-swe-agent, Terminal-Bench via terminus-2 with 3 repeats at 256K context, BFCL temperature 0.01.
- Context
- Where the card gives a per-benchmark protocol, it is included in the configuration. Other rows do not have a reported sample count in the table.
Humanity's Last Exam (text) (Show measurement, test conditions, and source)
- Source value
- 19.5
- Score
- 19.5
- Metric
- Humanity's Last Exam (text)
- Unit
- %
- Category
- source-specific
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- IFM model card
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Exact released K2 Horizon 7B anti_doom_loop_verified checkpoint after FTPO. IFM states before/after checkpoints use identical settings. The card supplies test-specific details: GPQA top_p=1 and 5 repeats, SciCode 3 repeats, SWE-bench via mini-swe-agent, Terminal-Bench via terminus-2 with 3 repeats at 256K context, BFCL temperature 0.01.
- Context
- Where the card gives a per-benchmark protocol, it is included in the configuration. Other rows do not have a reported sample count in the table.
IFEval (loose) (Show measurement, test conditions, and source)
- Source value
- 86.9
- Score
- 86.9
- Metric
- IFEval (loose)
- Unit
- %
- Category
- source-specific
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- IFM model card
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Exact released K2 Horizon 7B anti_doom_loop_verified checkpoint after FTPO. IFM states before/after checkpoints use identical settings. The card supplies test-specific details: GPQA top_p=1 and 5 repeats, SciCode 3 repeats, SWE-bench via mini-swe-agent, Terminal-Bench via terminus-2 with 3 repeats at 256K context, BFCL temperature 0.01.
- Context
- Where the card gives a per-benchmark protocol, it is included in the configuration. Other rows do not have a reported sample count in the table.
IFBench (loose) (Show measurement, test conditions, and source)
- Source value
- 50.7
- Score
- 50.7
- Metric
- IFBench (loose)
- Unit
- %
- Category
- source-specific
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- IFM model card
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Exact released K2 Horizon 7B anti_doom_loop_verified checkpoint after FTPO. IFM states before/after checkpoints use identical settings. The card supplies test-specific details: GPQA top_p=1 and 5 repeats, SciCode 3 repeats, SWE-bench via mini-swe-agent, Terminal-Bench via terminus-2 with 3 repeats at 256K context, BFCL temperature 0.01.
- Context
- Where the card gives a per-benchmark protocol, it is included in the configuration. Other rows do not have a reported sample count in the table.
LiveCodeBench v6 (Show measurement, test conditions, and source)
- Source value
- 72.7
- Score
- 72.7
- Metric
- LiveCodeBench v6
- Unit
- %
- Category
- source-specific
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- IFM model card
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Exact released K2 Horizon 7B anti_doom_loop_verified checkpoint after FTPO. IFM states before/after checkpoints use identical settings. The card supplies test-specific details: GPQA top_p=1 and 5 repeats, SciCode 3 repeats, SWE-bench via mini-swe-agent, Terminal-Bench via terminus-2 with 3 repeats at 256K context, BFCL temperature 0.01.
- Context
- Where the card gives a per-benchmark protocol, it is included in the configuration. Other rows do not have a reported sample count in the table.
OJBench (Show measurement, test conditions, and source)
- Source value
- 29.4
- Score
- 29.4
- Metric
- OJBench
- Unit
- %
- Category
- source-specific
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- IFM model card
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Exact released K2 Horizon 7B anti_doom_loop_verified checkpoint after FTPO. IFM states before/after checkpoints use identical settings. The card supplies test-specific details: GPQA top_p=1 and 5 repeats, SciCode 3 repeats, SWE-bench via mini-swe-agent, Terminal-Bench via terminus-2 with 3 repeats at 256K context, BFCL temperature 0.01.
- Context
- Where the card gives a per-benchmark protocol, it is included in the configuration. Other rows do not have a reported sample count in the table.
SciCode (Show measurement, test conditions, and source)
- Source value
- 33.6
- Score
- 33.6
- Metric
- SciCode
- Unit
- %
- Category
- source-specific
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- IFM model card
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Exact released K2 Horizon 7B anti_doom_loop_verified checkpoint after FTPO. IFM states before/after checkpoints use identical settings. The card supplies test-specific details: GPQA top_p=1 and 5 repeats, SciCode 3 repeats, SWE-bench via mini-swe-agent, Terminal-Bench via terminus-2 with 3 repeats at 256K context, BFCL temperature 0.01.
- Context
- Where the card gives a per-benchmark protocol, it is included in the configuration. Other rows do not have a reported sample count in the table.
SWE-bench Verified (Show measurement, test conditions, and source)
- Source value
- 72.4
- Score
- 72.4
- Metric
- SWE-bench Verified
- Unit
- %
- Category
- source-specific
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- IFM model card
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Vendor-reported result. Exact released K2 Horizon 7B anti_doom_loop_verified checkpoint after FTPO. IFM states before/after checkpoints use identical settings. The card supplies test-specific details: GPQA top_p=1 and 5 repeats, SciCode 3 repeats, SWE-bench via mini-swe-agent, Terminal-Bench via terminus-2 with 3 repeats at 256K context, BFCL temperature 0.01.
- Context
- Where the card gives a per-benchmark protocol, it is included in the configuration. Other rows do not have a reported sample count in the table.
Page 1 of 3
Sources and data date
| Type | Evidence and data date |
|---|---|
| Input | Documented modalities. Retrieved October 7, 2026. |
| Output | Documented modalities. Retrieved October 7, 2026. |
| Research status | Research date October 7, 2026. 11 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Unresolved | full-results-checkpoint-identity Original wording The Full Results table differs on several tests from the separate FTPO table for the released checkpoint. The published card does not identify the differing table's exact variant. |
| Not published | per-run-protocol-and-sample-counts Original wording Repeat counts are published only for some tests in the anti-doom-loop section; undisclosed sample sizes remain unknown. |
| Unresolved | excluded-non-target-checkpoints Original wording The profile description and model card identify anti_doom_loop_verified after FTPO as the final 7B checkpoint. mid_4_final is a different intermediate weight. The additional Full Results table and historical 82% SWE-bench value do not establish the exact final checkpoint; IFM also says the 82% value was inflated by downloaded answers. These values were excluded. Only the explicitly labelled after-FTPO results remain. |
| Additional source | IFM K2 Horizon release announcement (retrieved October 7, 2026) |
| Additional source | IFM model card (retrieved October 7, 2026) |
| Additional source | Artificial Analysis model profile (retrieved October 7, 2026) |
| Additional source | Model metadata (retrieved October 7, 2026) |
| Additional source | Model metadata (retrieved October 7, 2026) |
| Additional source | Model metadata (retrieved October 7, 2026) |
| Additional source | Model metadata (retrieved October 7, 2026) |
| Additional source | Model metadata (retrieved October 7, 2026) |
| Additional source | Model metadata (retrieved October 7, 2026) |
| Additional source | Model metadata (retrieved October 7, 2026) |
| Additional source | Model metadata (retrieved October 7, 2026) |
| Additional source | K2 Horizon 3.7B model card · Editorial description reviewed October 7, 2026 |