Skip to content

deprcate even more stuff, cut compute cost by 60% #2406

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
functionstackx merged 7 commits into from
Jul 29, 2026
Merged

deprcate even more stuff, cut compute cost by 60% #2406

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
26a50ce
docs: announce 2026-08-03 deprecations and spec-decode-only AgentX po…
functionstackx Jul 29, 2026
7892166
docs: justify spec-decode-only policy with the golden AL curves [skip…
functionstackx Jul 29, 2026
d837876
docs: mark the synthetic-acceptance knob list as non-exhaustive [skip…
functionstackx Jul 29, 2026
3dd9c82
docs: cut GLM-5.2 AgentX non-MTP on 2026-08-03 [skip-sweep]
functionstackx Jul 29, 2026
b4cac66
docs: drop the GLM-5.2 MTP-prerequisite paragraph [skip-sweep]
functionstackx Jul 29, 2026
3798962
fix(docs): MiniMax-M3 uses EAGLE3, not MTP [skip-sweep]
functionstackx Jul 29, 2026
ca783fc
Update MODELS.md
functionstackx Jul 29, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
45 changes: 37 additions & 8 deletions MODELS.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,13 +6,41 @@ This document tracks every model benchmarked by InferenceX-e2e: when it was adde

## Deprecation Notice

- **Thursday, August 6, 2026** is the last day for the **Single-turn 8k1k** scenario on **Kimi-K2.5/2.6/2.7-Code** (`kimik2.5`); the scenario is deprecated for these models after that date. Rationale: the Agentic coding (AgentX) scenario already covers them with real-world traffic, and Kimi-K3 launched on July 27, 2026, so GPU cluster time shifts to the newer frontier model.
InferenceX-e2e runs on a fixed, limited pool of GPUs and is maintained by a small team. Every scenario, precision, and recipe variant we keep alive consumes cluster hours and maintainer attention that would otherwise go to new frontier models. The deprecations below free that capacity. Where a deprecation removes one arm of an A/B pair, we keep and publish the arm that wins on the Pareto frontier.

### Monday, August 3, 2026

**Monday, August 3, 2026** is the last day for the scenarios, precisions, and recipe variants listed below; they are deprecated after that date.

Scenario and precision retirements:

| Model | Deprecated | Remains |
|---|---|---|
| MiniMax-M3 (`minimaxm3`) | Single-turn 8k1k | Agentic coding |
| Kimi-K2.5/2.6/2.7-Code (`kimik2.5`) | Agentic coding | Single-turn 8k1k, until August 6, 2026 (see below) |
| Qwen3.5-397B-A17B (`qwen3.5`) | All **bf16** recipes, in every scenario, on NVIDIA and AMD | fp8 and fp4 recipes |

Speculative-decoding A/B retirements — in each pair below the spec-decode arm is the better Pareto frontier, so we stop running the non-spec-decode arm and publish only the spec-decode arm:

| Model | Deprecated arm | Published arm |
|---|---|---|
| DeepSeek-V4-Pro 1.6T (`dsv4`) | Agentic coding, non-MTP | Agentic coding, MTP |
| Qwen3.5-397B-A17B (`qwen3.5`) | Agentic coding, non-MTP | Agentic coding, MTP |
| MiniMax-M3 (`minimaxm3`) | Agentic coding, non-EAGLE3 | Agentic coding, EAGLE3 |
| GLM-5.2 (`glm5.2`) | Agentic coding, non-MTP | Agentic coding, MTP |
| Kimi-K3 (`kimik3`) | Agentic coding, non-DSpark — deprecated from day 0 | Agentic coding, DSpark |
Comment on lines +21 to +31

@claude claude Bot Jul 29, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟡 The spec-decode A/B retirement table and matrix rows claim 'published' arms (Kimi-K3 DSpark; MiniMax-M3/Qwen3.5 agentic-MTP) that don't exist in configs/nvidia-master.yaml or configs/amd-master.yaml — for kimik3 both existing recipes are non-DSpark (companion PR #2391 shipped it non-DSpark because DSpark 'doesn't work with Pipeline yet'), and for minimaxm3/qwen3.5 all agentic-coding recipes have no spec-decoding field (their only -mtp variants are under the single-turn 8k1k scenario). This is a pre-existing documentation gap made worse by this PR: after 2026-08-03 these three rows will point to scenarios with no backing config, contradicting the PR's own test-plan claim that arms were cross-checked against the master configs. Recommend either dropping these three rows from the retirement table until the DSpark/MTP-agentic configs actually land, or rewording them as forward-looking intent rather than an active 'published arm'.

Extended reasoning...

The PR adds a "Speculative-decoding A/B retirements" table (MODELS.md L21-30, mirrored in MODELS_zh.md) plus corresponding matrix rows for kimik3 (L52), minimaxm3 (L54), and qwen3.5 (L59) that name a "Published arm" going forward: DSpark for Kimi-K3, and "Agentic coding, MTP" for MiniMax-M3 and Qwen3.5. None of these three published arms currently exist in configs/nvidia-master.yaml or configs/amd-master.yaml.

For kimik3: the only two recipes in nvidia-master.yaml are kimik3-fp4-b300-vllm-agentic (no spec-decoding field, defaults to none) and kimik3-fp4-b200-dynamo-vllm-agentic (explicit spec-decoding: none). There is no DSpark spec-decode config anywhere in either master file — the only "DSpark" string in the whole configs tree is an unrelated architecture comment about the vLLM plugin's MLA/KDA attention kernels. The companion PR #2391 (merged immediately before this one, commit 0decc69) shipped kimik3 non-DSpark specifically because "it doesn't work with Pipeline yet," with the DSpark work explicitly deferred to a follow-up. So this PR's own docs mark the only currently-runnable kimik3 arm as "deprecated from day 0" and mark the nonexistent arm as "Active"/"Published."

For minimaxm3 and qwen3.5: every agentic-coding recipe in both master configs (minimaxm3-fp8-{h100,h200}-vllm-agentic, amd -mi300x/-mi325x/-mi355x-vllm-agentic; qwen3.5-fp8-{b200,b300-hicache,h100}-sglang-agentic, amd -mi355x-sglang-agentic{,-hicache}) has no spec-decoding field. Their only -mtp recipe variants (e.g. minimaxm3-fp8-b200-vllm-mtp, qwen3.5-fp8-b200-sglang-mtp) live under the single-turn 8k1k scenario, not agentic-coding. A grep for agentic.*mtp across both master files returns only dsv4 and kimik2.5 — confirming dsv4's claimed arm in the same table is real, but minimaxm3/qwen3.5's is not.

The PR's own test plan states "All model prefixes, recipe variants, and spec-decode arms cross-checked against configs/nvidia-master.yaml and configs/amd-master.yaml" — this claim doesn't hold for these three rows. Concretely: (1) grep configs/nvidia-master.yaml and configs/amd-master.yaml for kimik3 → 2 hits, both spec-decoding: none or field-absent-defaults-to-none; (2) grep for dspark (case-insensitive) anywhere in the configs tree → 1 hit, a comment about attention kernels, not a spec-decode value; (3) grep agentic.*mtp across both files → only dsv4-* and kimik2.5-* keys, never minimaxm3-* or qwen3.5-*. Since this is a docs-only PR with [skip-sweep], nothing breaks at runtime today, but after 2026-08-03 the matrix would present an "active" scenario for three models backed by no config at all, which is exactly the kind of drift MODELS.md exists to prevent.

Fix: either don't table these as "published arm" until the follow-up PRs land the actual DSpark/agentic-MTP configs (kimik3's follow-up is already flagged as pending in #2391; minimaxm3/qwen3.5 would need new agentic+spec-decode recipes added first), or reword these three rows to state the intent to only publish spec-decode arms going forward rather than asserting a currently-existing "published arm."


**Going forward we no longer benchmark non-spec-decode versus spec-decode as an A/B.** The non-spec-decode arm existed as a neutral baseline back when acceptance length wasn't standardized. That is now solved: [`golden_al_distribution/`](golden_al_distribution/) commits one golden acceptance-length curve per model, thinking mode, and draft length, measured on the SPEED-Bench `coding` category, and AgentX pins every submission to that curve through synthetic acceptance (vLLM `synthetic_acceptance_length`, SGLang `SGLANG_SIMULATE_ACC_LEN`, TensorRT-LLM `TLLM_SPEC_DECODE_FORCE_NUM_ACCEPTED_TOKENS`, etc). With a fair, engine-independent acceptance target in place, spec-decode results are directly comparable on their own and a separate non-spec-decode track is redundant. Agentic coding recipes are therefore run and published with speculative decoding enabled only — MTP, EAGLE/EAGLE3, DSpark, or whatever draft method the model ships — and the non-spec-decode arm is neither run nor published. New models are onboarded that way from day 0, as Kimi-K3 is.

### Thursday, August 6, 2026

**Thursday, August 6, 2026** is the last day for the **Single-turn 8k1k** scenario on **Kimi-K2.5/2.6/2.7-Code** (`kimik2.5`); the scenario is deprecated for these models after that date. Rationale: Kimi-K3 launched on July 27, 2026, so GPU cluster time shifts to the newer frontier model. Combined with the Agentic coding deprecation above, this leaves `kimik2.5` with no active scenario — the model is **fully retired after August 6, 2026**.

## Scenarios

| Scenario | ISL/OSL | Status |
|---|---|---|
| Agentic coding | Long Context, Multi Turn Realistic traffic trace replay with sub agents | Active — trace-replay agentic-coding benchmark (see [`benchmarks/agentic/`](benchmarks/agentic/)). Going forward, new models will likely be onboarded with agentic coding only. |
| Agentic coding | Long Context, Multi Turn Realistic traffic trace replay with sub agents | Active — trace-replay agentic-coding benchmark (see [`benchmarks/agentic/`](benchmarks/agentic/)). Going forward, new models will likely be onboarded with agentic coding only, and **with speculative decoding enabled only** — the non-spec-decode arm is not run or published (see [Deprecation Notice](#deprecation-notice)). |
| Single-turn 8k1k | 8192 / 1024 | Active — the primary fixed-sequence-length scenario. |
| Single-turn 1k1k | 1024 / 1024 | **Deprecated for all models** since 2026-07-17 ([#2263](https://github.com/SemiAnalysisAI/InferenceX/pull/2263)), to save GPU cluster time for higher-priority real-world agentic-coding benchmarks and new frontier models. Archived configs live in [`configs/deprecated/`](configs/deprecated/). |
| Single-turn 1k8k | 1024 / 8192 | **Deprecated for all models** since 2026-03-27 ([#911](https://github.com/SemiAnalysisAI/InferenceX/pull/911)), to save GPU cluster time for higher-priority real-world agentic-coding benchmarks and new frontier models. Configs were removed, not archived. |
Expand All @@ -22,14 +50,14 @@ This document tracks every model benchmarked by InferenceX-e2e: when it was adde
| Model architecture class | Prefix | Date added | Active scenarios | Deprecated scenarios |
|---|---|---|---|---|
| Qwen3.8 2.4T | `qwen3.8` | TBD | Agentic coding | |
| Kimi-K3 | `kimik3` | 2026-07-27 ([#2391](https://github.com/SemiAnalysisAI/InferenceX/pull/2391)) | Agentic coding | |
| GLM-5.2 | `glm5.2` | 2026-07-18 ([#2268](https://github.com/SemiAnalysisAI/InferenceX/pull/2268)) | Agentic coding | |
| MiniMax-M3 | `minimaxm3` | 2026-06-12 ([#1724](https://github.com/SemiAnalysisAI/InferenceX/pull/1724)) | Single-turn 8k1k, Agentic coding | Single-turn 1k1k |
| DeepSeek-V4-Pro | `dsv4` | 2026-04-24 ([#1130](https://github.com/SemiAnalysisAI/InferenceX/pull/1130)) | Single-turn 8k1k, Agentic coding | Single-turn 1k1k |
| Kimi-K3 | `kimik3` | 2026-07-27 ([#2391](https://github.com/SemiAnalysisAI/InferenceX/pull/2391)) | Agentic coding (DSpark only) | Agentic coding non-DSpark arm (deprecated from day 0) |
| GLM-5.2 | `glm5.2` | 2026-07-18 ([#2268](https://github.com/SemiAnalysisAI/InferenceX/pull/2268)) | Agentic coding (MTP only from 2026-08-03) | |
| MiniMax-M3 | `minimaxm3` | 2026-06-12 ([#1724](https://github.com/SemiAnalysisAI/InferenceX/pull/1724)) | Single-turn 8k1k (until 2026-08-03), Agentic coding (EAGLE3 only from 2026-08-03) | Single-turn 1k1k |
| DeepSeek-V4-Pro | `dsv4` | 2026-04-24 ([#1130](https://github.com/SemiAnalysisAI/InferenceX/pull/1130)) | Single-turn 8k1k, Agentic coding (MTP only from 2026-08-03) | Single-turn 1k1k |
| GLM-5 / GLM-5.1 | `glm5`, `glm5.1` | 2026-03-06 ([#762](https://github.com/SemiAnalysisAI/InferenceX/pull/762)); GLM-5.1 added 2026-04-21 ([#1098](https://github.com/SemiAnalysisAI/InferenceX/pull/1098)) | — (retired 2026-07-18, [#2276](https://github.com/SemiAnalysisAI/InferenceX/pull/2276)) | Single-turn 1k1k, Single-turn 1k8k (GLM-5 only), Single-turn 8k1k |
| MiniMax-M2.5/2.7 | `minimaxm2.5` | 2026-02-18 ([#755](https://github.com/SemiAnalysisAI/InferenceX/pull/755)) | — (retired 2026-06-20, [#1874](https://github.com/SemiAnalysisAI/InferenceX/pull/1874)) | Single-turn 1k1k, Single-turn 1k8k, Single-turn 8k1k |
| Kimi-K2.5/2.6/2.7-Code | `kimik2.5` | 2026-02-17 ([#734](https://github.com/SemiAnalysisAI/InferenceX/pull/734)) | Single-turn 8k1k, Agentic coding | Single-turn 1k1k, Single-turn 1k8k |
| Qwen3.5-397B-A17B | `qwen3.5` | 2026-02-16 ([#704](https://github.com/SemiAnalysisAI/InferenceX/pull/704)) | Single-turn 8k1k, Agentic coding | Single-turn 1k1k, Single-turn 1k8k |
| Kimi-K2.5/2.6/2.7-Code | `kimik2.5` | 2026-02-17 ([#734](https://github.com/SemiAnalysisAI/InferenceX/pull/734)) | Single-turn 8k1k (until 2026-08-06), Agentic coding (until 2026-08-03) — fully retired after 2026-08-06 | Single-turn 1k1k, Single-turn 1k8k |
| Qwen3.5-397B-A17B | `qwen3.5` | 2026-02-16 ([#704](https://github.com/SemiAnalysisAI/InferenceX/pull/704)) | Single-turn 8k1k, Agentic coding (MTP only from 2026-08-03); fp8/fp4 only — bf16 recipes retired 2026-08-03 | Single-turn 1k1k, Single-turn 1k8k |
| gpt-oss-120b | `gptoss` | 2025-09-09 | — (retired 2026-07-06, [#2101](https://github.com/SemiAnalysisAI/InferenceX/pull/2101)) | Single-turn 1k1k, Single-turn 1k8k, Single-turn 8k1k |
| DeepSeek-R1-0528 | `dsr1` | 2025-08-13 | Single-turn 8k1k | Single-turn 1k1k, Single-turn 1k8k |
| Llama-3.1-70B-Instruct | `llama70b` | 2025-08-12 | — (retired 2025-10-29, [#149](https://github.com/SemiAnalysisAI/InferenceX/pull/149)) | Single-turn 1k1k, Single-turn 1k8k, Single-turn 8k1k [^1] |
Expand All @@ -40,5 +68,6 @@ This document tracks every model benchmarked by InferenceX-e2e: when it was adde

- The `Prefix` column is the canonical `model-prefix` used in `configs/*-master.yaml` and by `generate_sweep_configs.py --model-prefix`.
- "Retired" means the model no longer has any active scenario. Retired models' configs (except `llama70b`) are archived under [`configs/deprecated/`](configs/deprecated/).
- Deprecating a precision (e.g. Qwen3.5 bf16) or one arm of an A/B pair (e.g. non-MTP) narrows a model's recipe coverage without retiring the model; the model stays listed as active as long as one scenario still runs.
- `dsr1` began as the DeepSeek-V3 workflow templates in the initial repo import and was switched to DeepSeek-R1 benchmarking on 2025-08-13 (renamed `dsv3` → `dsr1` on 2025-08-20).
- Adding a model? Follow "Adding a benchmark configuration" in [`AGENTS.md`](AGENTS.md) and add a row here (and in [`MODELS_zh.md`](MODELS_zh.md)) in the same PR.
45 changes: 37 additions & 8 deletions MODELS_zh.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,13 +6,41 @@

## 弃用公告

- **2026 年 8 月 6 日（星期四）**为 **Kimi-K2.5/2.6/2.7-Code**（`kimik2.5`）**单轮 8k1k** 场景的最后运行日，此后该场景对这些模型弃用。原因：智能体编码（AgentX）场景已以真实流量覆盖这些模型，且 Kimi-K3 已于 2026 年 7 月 27 日发布，GPU 集群时间将转向更新的前沿模型。
InferenceX-e2e 运行在数量固定且有限的 GPU 资源池上，并由一支小型团队维护。每保留一个场景、精度或配方变体，都会占用集群机时与维护人力，而这些资源本可投入到新的前沿模型上。以下弃用即为释放这部分产能。若某项弃用移除的是 A/B 对照中的一个分支，我们保留并发布在帕累托前沿（Pareto frontier）上更优的那个分支。

### 2026 年 8 月 3 日（星期一）

**2026 年 8 月 3 日（星期一）**为下列场景、精度与配方变体的最后运行日，此后即告弃用。

场景与精度下线：

| 模型 | 弃用内容 | 保留内容 |
|---|---|---|
| MiniMax-M3（`minimaxm3`） | 单轮 8k1k | 智能体编码 |
| Kimi-K2.5/2.6/2.7-Code（`kimik2.5`） | 智能体编码 | 单轮 8k1k，保留至 2026 年 8 月 6 日（见下文） |
| Qwen3.5-397B-A17B（`qwen3.5`） | 全部 **bf16** 配方，涵盖所有场景，NVIDIA 与 AMD 平台均在内 | fp8 与 fp4 配方 |

投机解码（speculative decoding）A/B 对照下线 —— 下列每一组对照中，启用投机解码的分支都处于更优的帕累托前沿，因此我们停止运行非投机解码分支，仅发布投机解码分支：

| 模型 | 弃用分支 | 发布分支 |
|---|---|---|
| DeepSeek-V4-Pro 1.6T（`dsv4`） | 智能体编码，非 MTP | 智能体编码，MTP |
| Qwen3.5-397B-A17B（`qwen3.5`） | 智能体编码，非 MTP | 智能体编码，MTP |
| MiniMax-M3（`minimaxm3`） | 智能体编码，非 EAGLE3 | 智能体编码，EAGLE3 |
| GLM-5.2（`glm5.2`） | 智能体编码，非 MTP | 智能体编码，MTP |
| Kimi-K3（`kimik3`） | 智能体编码，非 DSpark —— 自第 0 天（day 0）起即弃用 | 智能体编码，DSpark |

**今后我们不再以 A/B 对照的方式基准测试「非投机解码 vs 投机解码」。**当初保留非投机解码分支，是把它当作中立基线：那时接受长度（AL）完全取决于提交方草稿头（draft head）的实际水平，导致各家投机解码数据之间无法横向比较。这一问题现已解决：[`golden_al_distribution/`](golden_al_distribution/) 为每个模型、thinking 模式与草稿长度各提交了一条黄金 AL 曲线，均在 SPEED-Bench `coding` 类别上测得；AgentX 通过合成接受（synthetic acceptance）将所有提交锁定到该曲线（vLLM 用 `synthetic_acceptance_length`，SGLang 用 `SGLANG_SIMULATE_ACC_LEN`，TensorRT-LLM 用 `TLLM_SPEC_DECODE_FORCE_NUM_ACCEPTED_TOKENS`，等等）。既然已有公平且与引擎无关的接受目标，投机解码结果本身即可直接横向比较，单独保留一条非投机解码赛道已属冗余。因此，智能体编码配方一律仅在启用投机解码的条件下运行与发布 —— 具体为 MTP、EAGLE/EAGLE3、DSpark，或该模型自带的任何草稿方法 —— 非投机解码分支既不运行也不发布。新模型自第 0 天起即按此方式接入，Kimi-K3 即为一例。

### 2026 年 8 月 6 日（星期四）

**2026 年 8 月 6 日（星期四）**为 **Kimi-K2.5/2.6/2.7-Code**（`kimik2.5`）**单轮 8k1k** 场景的最后运行日，此后该场景对这些模型弃用。原因：Kimi-K3 已于 2026 年 7 月 27 日发布，GPU 集群时间将转向更新的前沿模型。叠加上文的智能体编码弃用，`kimik2.5` 将不再有任何启用场景 —— 该模型将于 **2026 年 8 月 6 日后完全退役**。

## 场景

| 场景 | ISL/OSL | 状态 |
|---|---|---|
| 智能体编码（agentic coding） | 长上下文、多轮真实流量的轨迹回放，含子智能体（sub agents） | 启用 — 基于轨迹回放的智能体编码基准测试（见 [`benchmarks/agentic/`](benchmarks/agentic/)）。今后新模型预计将仅以智能体编码场景接入。 |
| 智能体编码（agentic coding） | 长上下文、多轮真实流量的轨迹回放，含子智能体（sub agents） | 启用 — 基于轨迹回放的智能体编码基准测试（见 [`benchmarks/agentic/`](benchmarks/agentic/)）。今后新模型预计将仅以智能体编码场景接入，且**仅在启用投机解码的条件下**运行 —— 非投机解码分支不运行也不发布（见[弃用公告](#弃用公告)）。 |
| 单轮 8k1k | 8192 / 1024 | 启用 — 当前主要的固定序列长度（fixed-seq-len）场景。 |
| 单轮 1k1k | 1024 / 1024 | **对所有模型均已弃用**，自 2026-07-17 起（[#2263](https://github.com/SemiAnalysisAI/InferenceX/pull/2263)），以便将 GPU 集群时间留给优先级更高的真实场景智能体编码基准测试与新的前沿模型。归档配置位于 [`configs/deprecated/`](configs/deprecated/)。 |
| 单轮 1k8k | 1024 / 8192 | **对所有模型均已弃用**，自 2026-03-27 起（[#911](https://github.com/SemiAnalysisAI/InferenceX/pull/911)），以便将 GPU 集群时间留给优先级更高的真实场景智能体编码基准测试与新的前沿模型。相关配置已删除，未归档。 |
Expand All @@ -22,14 +50,14 @@
| 模型架构类别 | 前缀 | 加入日期 | 启用场景 | 已弃用场景 |
|---|---|---|---|---|
| Qwen3.8 2.4T | `qwen3.8` | 待定 | 智能体编码 | |
| Kimi-K3 | `kimik3` | 2026-07-27 ([#2391](https://github.com/SemiAnalysisAI/InferenceX/pull/2391)) | 智能体编码 | |
| GLM-5.2 | `glm5.2` | 2026-07-18（[#2268](https://github.com/SemiAnalysisAI/InferenceX/pull/2268)） | 智能体编码 | |
| MiniMax-M3 | `minimaxm3` | 2026-06-12（[#1724](https://github.com/SemiAnalysisAI/InferenceX/pull/1724)） | 单轮 8k1k、智能体编码 | 单轮 1k1k |
| DeepSeek-V4-Pro | `dsv4` | 2026-04-24（[#1130](https://github.com/SemiAnalysisAI/InferenceX/pull/1130)） | 单轮 8k1k、智能体编码 | 单轮 1k1k |
| Kimi-K3 | `kimik3` | 2026-07-27 ([#2391](https://github.com/SemiAnalysisAI/InferenceX/pull/2391)) | 智能体编码（仅 DSpark） | 智能体编码非 DSpark 分支（自第 0 天起弃用） |
| GLM-5.2 | `glm5.2` | 2026-07-18（[#2268](https://github.com/SemiAnalysisAI/InferenceX/pull/2268)） | 智能体编码（自 2026-08-03 起仅 MTP） | |
| MiniMax-M3 | `minimaxm3` | 2026-06-12（[#1724](https://github.com/SemiAnalysisAI/InferenceX/pull/1724)） | 单轮 8k1k（至 2026-08-03）、智能体编码（自 2026-08-03 起仅 EAGLE3） | 单轮 1k1k |
| DeepSeek-V4-Pro | `dsv4` | 2026-04-24（[#1130](https://github.com/SemiAnalysisAI/InferenceX/pull/1130)） | 单轮 8k1k、智能体编码（自 2026-08-03 起仅 MTP） | 单轮 1k1k |
| GLM-5 / GLM-5.1 | `glm5`、`glm5.1` | 2026-03-06（[#762](https://github.com/SemiAnalysisAI/InferenceX/pull/762)）；GLM-5.1 于 2026-04-21 加入（[#1098](https://github.com/SemiAnalysisAI/InferenceX/pull/1098)） | —（2026-07-18 退役，[#2276](https://github.com/SemiAnalysisAI/InferenceX/pull/2276)） | 单轮 1k1k、单轮 1k8k（仅 GLM-5）、单轮 8k1k |
| MiniMax-M2.5/2.7 | `minimaxm2.5` | 2026-02-18（[#755](https://github.com/SemiAnalysisAI/InferenceX/pull/755)） | —（2026-06-20 退役，[#1874](https://github.com/SemiAnalysisAI/InferenceX/pull/1874)） | 单轮 1k1k、单轮 1k8k、单轮 8k1k |
| Kimi-K2.5/2.6/2.7-Code | `kimik2.5` | 2026-02-17（[#734](https://github.com/SemiAnalysisAI/InferenceX/pull/734)） | 单轮 8k1k、智能体编码 | 单轮 1k1k、单轮 1k8k |
| Qwen3.5-397B-A17B | `qwen3.5` | 2026-02-16（[#704](https://github.com/SemiAnalysisAI/InferenceX/pull/704)） | 单轮 8k1k、智能体编码 | 单轮 1k1k、单轮 1k8k |
| Kimi-K2.5/2.6/2.7-Code | `kimik2.5` | 2026-02-17（[#734](https://github.com/SemiAnalysisAI/InferenceX/pull/734)） | 单轮 8k1k（至 2026-08-06）、智能体编码（至 2026-08-03）—— 2026-08-06 后完全退役 | 单轮 1k1k、单轮 1k8k |
| Qwen3.5-397B-A17B | `qwen3.5` | 2026-02-16（[#704](https://github.com/SemiAnalysisAI/InferenceX/pull/704)） | 单轮 8k1k、智能体编码（自 2026-08-03 起仅 MTP）；仅 fp8/fp4 —— bf16 配方于 2026-08-03 下线 | 单轮 1k1k、单轮 1k8k |
| gpt-oss-120b | `gptoss` | 2025-09-09 | —（2026-07-06 退役，[#2101](https://github.com/SemiAnalysisAI/InferenceX/pull/2101)） | 单轮 1k1k、单轮 1k8k、单轮 8k1k |
| DeepSeek-R1-0528 | `dsr1` | 2025-08-13 | 单轮 8k1k | 单轮 1k1k、单轮 1k8k |
| Llama-3.1-70B-Instruct | `llama70b` | 2025-08-12 | —（2025-10-29 退役，[#149](https://github.com/SemiAnalysisAI/InferenceX/pull/149)） | 单轮 1k1k、单轮 1k8k、单轮 8k1k [^1] |
Expand All @@ -40,5 +68,6 @@

- 「前缀」列为 `configs/*-master.yaml` 中的规范 `model-prefix`，同时用于 `generate_sweep_configs.py --model-prefix`。
- 「退役」指该模型已无任何启用场景。退役模型的配置（`llama70b` 除外）归档于 [`configs/deprecated/`](configs/deprecated/)。
- 弃用某一精度（如 Qwen3.5 bf16）或 A/B 对照中的某一分支（如非 MTP），只是收窄该模型的配方覆盖范围，并不等于模型退役；只要仍有一个场景在运行，该模型即继续列为启用状态。
- `dsr1` 最初以 DeepSeek-V3 workflow 模板的形式随仓库首次导入，2025-08-13 切换为 DeepSeek-R1 基准测试（2025-08-20 将 `dsv3` 重命名为 `dsr1`）。
- 新增模型时，请按 [`AGENTS.md`](AGENTS.md) 中「Adding a benchmark configuration」的流程操作，并在同一 PR 中同时更新本文件与 [`MODELS.md`](MODELS.md) 的表格。